Podcasts about Andrew Gelman

Play Episode Listen Later May 16, 2024 76:47 Transcription Available

Proudly sponsored by PyMC Labs, the Bayesian Consultancy. Book a call, or get in touch!My Intuitive Bayes Online Courses1:1 Mentorship with meIf there is one guest I don't need to introduce, it's mister Andrew Gelman. So… I won't! I will refer you back to his two previous appearances on the show though, because learning from Andrew is always a pleasure. So go ahead and listen to episodes 20 and 27.In this episode, Andrew and I discuss his new book, Active Statistics, which focuses on teaching and learning statistics through active student participation. Like this episode, the book is divided into three parts: 1) The ideas of statistics, regression, and causal inference; 2) The value of storytelling to make statistical concepts more relatable and interesting; 3) The importance of teaching statistics in an active learning environment, where students are engaged in problem-solving and discussion.And Andrew is so active and knowledgeable that we of course touched on a variety of other topics — but for that, you'll have to listen ;)Our theme music is « Good Bayesian », by Baba Brinkman (feat MC Lars and Mega Ran). Check out his awesome work at https://bababrinkman.com/ !Thank you to my Patrons for making this episode possible!Yusuke Saito, Avi Bryant, Ero Carrera, Giuliano Cruz, Tim Gasser, James Wade, Tradd Salvo, William Benton, James Ahloy, Robin Taylor,, Chad Scherrer, Zwelithini Tunyiswa, Bertrand Wilden, James Thompson, Stephen Oates, Gian Luca Di Tanna, Jack Wells, Matthew Maldonado, Ian Costley, Ally Salim, Larry Gill, Ian Moran, Paul Oreto, Colin Caprani, Colin Carroll, Nathaniel Burbank, Michael Osthege, Rémi Louf, Clive Edelsten, Henri Wallen, Hugo Botha, Vinh Nguyen, Marcin Elantkowski, Adam C. Smith, Will Kurt, Andrew Moskowitz, Hector Munoz, Marco Gorelli, Simon Kessell, Bradley Rode, Patrick Kelley, Rick Anderson, Casper de Bruin, Philippe Labonde, Michael Hankin, Cameron Smith, Tomáš Frýda, Ryan Wesslen, Andreas Netti, Riley King, Yoshiyuki Hamajima, Sven De Maeyer, Michael DeCrescenzo, Fergal M, Mason Yahr, Naoya Kanai, Steven Rowland, Aubrey Clayton, Jeannine Sue, Omri Har Shemesh, Scott Anthony Robson, Robert Yolken, Or Duek, Pavel Dusek, Paul Cox, Andreas Kröpelin, Raphaël R, Nicolas Rode, Gabriel Stechschulte, Arkady, Kurt TeKolste, Gergely Juhasz, Marcus Nölke, Maggi Mackintosh, Grant Pezzolesi, Avram Aelony, Joshua Meehl, Javier Sabio, Kristian Higgins, Alex Jones, Gregorio Aguilar, Matt Rosinski, Bart Trudeau, Luis Fonseca, Dante Gates, Matt Niccolls, Maksim Kuznecov, Michael Thomas, Luke Gorrie, Cory Kiser, Julio, Edvin Saveljev, Frederick Ayala, Jeffrey Powell, Gal Kampel, Adan Romero, Will Geary and Blake Walters.Visit https://www.patreon.com/learnbayesstats to unlock exclusive Bayesian swag ;)Takeaways:- Active learning is essential for teaching and learning statistics.- Storytelling can make...

#90, Demystifying MCMC & Variational Inference, with Charles Margossian

relationships law reciprocity building a team meaningful relationships andrew gelman

Play Episode Listen Later Sep 6, 2023 97:36 Transcription Available

Proudly sponsored by PyMC Labs, the Bayesian Consultancy. Book a call, or get in touch!My Intuitive Bayes Online Courses1:1 Mentorship with meWhat's the difference between MCMC and Variational Inference (VI)? Why is MCMC called an approximate method? When should we use VI instead of MCMC?These are some of the captivating (and practical) questions we'll tackle in this episode. I had the chance to interview Charles Margossian, a research fellow in computational mathematics at the Flatiron Institute, and a core developer of the Stan software.Charles was born and raised in Paris, and then moved to the US to pursue a bachelor's degree in physics at Yale university. After graduating, he worked for two years in biotech, and went on to do a PhD in statistics at Columbia University with someone named… Andrew Gelman — you may have heard of him.Charles is also specialized in pharmacometrics and epidemiology, so we also talked about some practical applications of Bayesian methods and algorithms in these fascinating fields.Oh, and Charles' life doesn't only revolve around computers: he practices ballroom dancing and pickup soccer, and used to do improvised musical comedy!Our theme music is « Good Bayesian », by Baba Brinkman (feat MC Lars and Mega Ran). Check out his awesome work at https://bababrinkman.com/ !Thank you to my Patrons for making this episode possible!Yusuke Saito, Avi Bryant, Ero Carrera, Giuliano Cruz, Tim Gasser, James Wade, Tradd Salvo, William Benton, James Ahloy, Robin Taylor,, Chad Scherrer, Zwelithini Tunyiswa, Bertrand Wilden, James Thompson, Stephen Oates, Gian Luca Di Tanna, Jack Wells, Matthew Maldonado, Ian Costley, Ally Salim, Larry Gill, Ian Moran, Paul Oreto, Colin Caprani, Colin Carroll, Nathaniel Burbank, Michael Osthege, Rémi Louf, Clive Edelsten, Henri Wallen, Hugo Botha, Vinh Nguyen, Raul Maldonado, Marcin Elantkowski, Adam C. Smith, Will Kurt, Andrew Moskowitz, Hector Munoz, Marco Gorelli, Simon Kessell, Bradley Rode, Patrick Kelley, Rick Anderson, Casper de Bruin, Philippe Labonde, Michael Hankin, Cameron Smith, Tomáš Frýda, Ryan Wesslen, Andreas Netti, Riley King, Yoshiyuki Hamajima, Sven De Maeyer, Michael DeCrescenzo, Fergal M, Mason Yahr, Naoya Kanai, Steven Rowland, Aubrey Clayton, Jeannine Sue, Omri Har Shemesh, Scott Anthony Robson, Robert Yolken, Or Duek, Pavel Dusek, Paul Cox, Trey Causey, Andreas Kröpelin, Raphaël R, Nicolas Rode, Gabriel Stechschulte, Arkady, Kurt TeKolste, Gergely Juhasz, Marcus Nölke, Maggi Mackintosh, Grant Pezzolesi, Avram Aelony, Joshua Meehl, Javier Sabio, Kristian Higgins, Alex Jones, Gregorio Aguilar and Matt Rosinski.Visit https://www.patreon.com/learnbayesstats to unlock exclusive Bayesian swag ;)Links from the show:Charles' website: https://charlesm93.github.io/Charles on Twitter: https://twitter.com/charlesm993Charles on GitHub:

phd columbia university mentorship yale alex jones demystifying casper github rapha proudly bruin bayesian cameron smith inference mega ran james thompson arkady james wade paul cox mc lars mcmc baba brinkman rick anderson robin taylor andrew gelman patrick kelley andreas kr aubrey clayton trey causey adam c smith

Episode 34: September 2023

JHLT: The Podcast

Play Episode Listen Later Sep 6, 2023 25:01

In this special issue of JHLT: The Podcast, the JHLT Digital Media Editors explore just one study—and devote the second half of the episode as a tribute to recently retired Editor-in-Chief, Dr. Daniel R. Goldstein. Digital Media Editor Van-Khue Ton, MD, heart failure and transplant cardiologist at Massachusetts General Hospital, hosts this episode. First, hear from senior author William F. Parker, MD, MS, PhD, on his team's study “Association of high-priority exceptions with waitlist mortality among heart transplant candidates.” Dr. Parker is a pulmonary and critical care physician, health services researcher, and clinical medical ethicist, and he runs a R01 funded lab focusing on the allocation of scarce healthcare resources. In the study, Dr. Parker and colleagues set out to examine the Scientific Registry of Transplant Recipients (SRTR) to compare medical urgency of heart transplant patients listed with exception vs. those listed according to standard guidelines. The study's main finding: after controlling for status as a time-varying covariate, candidates with an exception had a 45% lower hazard of waitlist mortality compared with standard criteria candidates. The Digital Media Editors want to know all the details and talk with Dr. Parker about the wait-list mortality and post-transplant survival of status 1, 2, 3, and 4 candidates, plus what next steps could be in ensuring a fair allocation system. The episode's special tribute to Dr. Daniel R. Goldstein features Michelle Kittleson, MD, PhD, Interim Editor-in-Chief of JHLT; Andrew Gelman, PhD, Deputy Editor at JHLT; Andrew Fisher, FRCP, PhD, past president of ISHLT and Chair of the Publications Oversight Committee; Daniel Tyrrell, PhD, a former post-doc of Dr. Goldstein's; and Judy Chen, PhD, a former immunology PhD student in Dr. Goldstein's lab. These heartfelt tributes to Dr. Goldstein are worth a listen—and we thank Dr. Goldstein for his vision and leadership at the Journal. Follow along at www.jhltonline.org/current, or, if you're an ISHLT member, log in at ishlt.org/journal-of-heart-lung-transplantation. Don't already get the Journal and want to read along? Join the International Society of Heart and Lung Transplantation at www.ishlt.org for a free subscription, or subscribe today at www.jhltonline.org.

phd ms heart medicine journal md medical pediatrics goldstein international society massachusetts general hospital medical research deputy editor frcp andrew fisher immunosuppression heart transplantation andrew gelman judy chen

How to Create More Meaningful Relationships in Business

The Jasmine Star Show

Play Episode Listen Later Aug 22, 2023 51:33 Transcription Available

Wanna know how to become someone that people WANT to work with?Build meaningful relationships. As a business owner, building and sustaining meaningful relationships will help you build trust and credibility, which can lead to more referrals, collaborations, and customers and clients.You might be thinking, ‘Ok, I've heard this before, but HOW… how do I create and build these relationships with people?'Lean in, this episode was made for you. I'm diving deep into four tried and true ways to build and maintain connections. Through intentional listening, being a connector, leading by example, and showing gratitude you can make people feel significant, seen, and validated…and build the type of relationships that make others love working with you. Click play to hear all of this and…[00:04:27] How to intentionally listen to your audience through content and conversations.[00:08:17] How to create a safe space for your team by encouraging open communication and honest and open dialogue.[00:02:21] The importance of seeking feedback from customers and clients for future content and offers.[00:09:25] The importance of openly and safely sharing perspectives in a team meeting to make informed decisions.[00:10:22] Why my team and I share pre-meeting notes and action items in advance and how it improves productivity and ensures everyone has a chance to be heard.[00:11:26] The importance of active listening, maintaining eye contact, nodding, and using appropriate facial expressions and body language to convey engagement.[00:19:17] Ways you can follow up with leads by creating content based on their ideas.[00:20:09] How connecting community members can help them learn and scale their businesses together.[00:21:01] The Law of Reciprocity and how connecting with others can build trust and lead to new relationships.[00:29:41] How to lead by example by demonstrating core values, investing in personal growth, and showing resilience and adaptability.[00:36:18] The importance of following through on promises, addressing issues for all customers, and maintaining trust and reliability.[00:38:15] The importance of answering messages promptly to maintain trust.[00:39:11] The significance of expressing gratitude to community members through sharing their content and sending personalized thank you messages.[00:45:59] How showing gratitude to customers and clients strengthens emotional connections and increases lifetime value. [00:48:54] The importance of publicly acknowledging clients' support and achievements on social media to make them feel valued and appreciated (and tips on how to do it!).For complete show notes, visit: http://jasminestar.com/podcast/episode361Sources:Cision PR Newswire, Motista, https://www.prnewswire.com/news-releases/new-retail-study-shows-marketers-under-leverage-emotional-connection-300720049.html The New York Times, Andrew Gelman, https://www.nytimes.com/2013/02/19/science/the-average-american-knows-how-many-people.html

6-26-23 INTERVIEW Andrew Gelman Professor at Columbia University on the Scandal at Harvard

The Ross Kaminsky Show

Play Episode Listen Later Jun 26, 2023 11:16

professor harvard scandals columbia university andrew gelman

LW - Are Bayesian methods guaranteed to overfit? by Ege Erdil

The Nonlinear Library

Play Episode Listen Later Jun 17, 2023 4:30

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Are Bayesian methods guaranteed to overfit?, published by Ege Erdil on June 17, 2023 on LessWrong. Yuling Yao argues that Bayesian models are guaranteed to overfit. He summarizes his point as follows: I have a different view. Bayesian model does overfit. Moreover, Bayes is guaranteed to overfit, regardless of the model (being correct or wrong) or the prior ( “strong” or uninformative). Moreover, Bayes is guaranteed to overfit on every realization of training data, not just in expectation. Moreover, Bayes is guaranteed to overfit on every single point of the training data, not just in the summation. He uses the following definition of "overfitting": a model "overfits" some data if its out-of-sample log loss exceeds its within-sample log loss. Interpreted in a different way, this is equivalent to saying that the model assigns higher probability to a data point after updating on it than before. Andrew Gelman makes the point that any proper fitting procedure whatsoever has this property, and alternative methods "overfit" more than ideal Bayesian methods. I think the proper way to interpret the results is not that Bayesian methods are guaranteed to overfit but that the definition of "overfitting" used by Yuling Yao, while intuitively plausible at first glance, is actually poor. Still, proving the fact that Bayesian methods indeed must "overfit" in his sense is an interesting exercise. I tried understanding his derivation of this and gave up - I present an original derivation of the same fact below that I hope is clearer. Derivation Suppose we have a model parametrized by parameters θ and the probability of seeing some data y according to our model is P(y|θ). Now, suppose we draw n independent samples y1,y2,.,yn. Denote this whole data vector by y, and denote the data vector with the ith sample omitted by y−i. Under Bayesian inference, the within-sample probability of observing the value yi in the next sample we draw is P(yn+1=yi|y)=∫θP(θ|y)P(yi|θ)dθ On the other hand, Bayes says that P(θ|y)=P(θ|y−i,yi)=P(θ|y−i)P(yi|y−i,θ)P(yi|y−i)=P(θ|y−i)P(yi|θ)P(yi|y−i) Plugging in gives P(yn+1=yi|y)=∫θP(θ|y−i)P(yi|θ)2P(yi|y−i)dθ or P(yn+1=yi|y)P(yi|y−i)=Eθ∼P(θ|y−i)[P(yi|θ)2] We can decompose the expectation of the squared probability on the right hand side using the definition of variance as follows: P(yn+1=yi|y)P(yi|y−i)=Eθ∼P(θ|y−i)[P(yi|θ)]2+varθ∼P(θ|y−i)(P(yi|θ))=P(yi|y−i)2+varθ∼P(θ|y−i)(P(yi|θ)) where I've used the fact that Eθ∼P(θ|y−i)[P(yi|θ)]=∫θP(θ|y−i)P(yi|θ)dθ=P(yi|y−i) to get rid of the expectation. The variance term on the right hand side is nonnegative by definition as it's a variance, and it's strictly positive as long as there's any uncertainty in our beliefs about θ after seeing the data y−i that would influence our probability estimate of observing yi next. This will be the case in almost all nondegenerate situations, and if so, we obtain the strict inequality P(yn+1=yi|y)>P(yi|y−i) What does this mean? The finding is intuitively obvious, but poses some challenges to formally defining the notion of overfitting. This is essentially because the ideal amount of fitting for a model to do on some data is nonzero, and overfitting should be "exceeding" this level of ideal fitting. In practice, though, it's difficult to know what is the "appropriate" amount of fitting for a model to be doing. Bayesian inference is ideal if the true model is within the class of models under consideration, but it might fail in unexpected ways if it's not, which is almost always the case in practice. I think the lesson to draw from this is that overfitting is a relative concept and claiming that a particular method "overfits" the data doesn't make too much sense without a point of reference in mind. If people have alternative ways of trying to construct an absolute notion of overfitting with the a...

speech ea guaranteed plugging bayesian bayes interpreted rationalist 2p lesswrong andrew gelman bayesian methods

LW - Are Bayesian methods guaranteed to overfit? by Ege Erdil

The Nonlinear Library: LessWrong

Play Episode Listen Later Jun 17, 2023 4:30

Link to original articleWelcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Are Bayesian methods guaranteed to overfit?, published by Ege Erdil on June 17, 2023 on LessWrong. Yuling Yao argues that Bayesian models are guaranteed to overfit. He summarizes his point as follows: I have a different view. Bayesian model does overfit. Moreover, Bayes is guaranteed to overfit, regardless of the model (being correct or wrong) or the prior ( “strong” or uninformative). Moreover, Bayes is guaranteed to overfit on every realization of training data, not just in expectation. Moreover, Bayes is guaranteed to overfit on every single point of the training data, not just in the summation. He uses the following definition of "overfitting": a model "overfits" some data if its out-of-sample log loss exceeds its within-sample log loss. Interpreted in a different way, this is equivalent to saying that the model assigns higher probability to a data point after updating on it than before. Andrew Gelman makes the point that any proper fitting procedure whatsoever has this property, and alternative methods "overfit" more than ideal Bayesian methods. I think the proper way to interpret the results is not that Bayesian methods are guaranteed to overfit but that the definition of "overfitting" used by Yuling Yao, while intuitively plausible at first glance, is actually poor. Still, proving the fact that Bayesian methods indeed must "overfit" in his sense is an interesting exercise. I tried understanding his derivation of this and gave up - I present an original derivation of the same fact below that I hope is clearer. Derivation Suppose we have a model parametrized by parameters θ and the probability of seeing some data y according to our model is P(y|θ). Now, suppose we draw n independent samples y1,y2,.,yn. Denote this whole data vector by y, and denote the data vector with the ith sample omitted by y−i. Under Bayesian inference, the within-sample probability of observing the value yi in the next sample we draw is P(yn+1=yi|y)=∫θP(θ|y)P(yi|θ)dθ On the other hand, Bayes says that P(θ|y)=P(θ|y−i,yi)=P(θ|y−i)P(yi|y−i,θ)P(yi|y−i)=P(θ|y−i)P(yi|θ)P(yi|y−i) Plugging in gives P(yn+1=yi|y)=∫θP(θ|y−i)P(yi|θ)2P(yi|y−i)dθ or P(yn+1=yi|y)P(yi|y−i)=Eθ∼P(θ|y−i)[P(yi|θ)2] We can decompose the expectation of the squared probability on the right hand side using the definition of variance as follows: P(yn+1=yi|y)P(yi|y−i)=Eθ∼P(θ|y−i)[P(yi|θ)]2+varθ∼P(θ|y−i)(P(yi|θ))=P(yi|y−i)2+varθ∼P(θ|y−i)(P(yi|θ)) where I've used the fact that Eθ∼P(θ|y−i)[P(yi|θ)]=∫θP(θ|y−i)P(yi|θ)dθ=P(yi|y−i) to get rid of the expectation. The variance term on the right hand side is nonnegative by definition as it's a variance, and it's strictly positive as long as there's any uncertainty in our beliefs about θ after seeing the data y−i that would influence our probability estimate of observing yi next. This will be the case in almost all nondegenerate situations, and if so, we obtain the strict inequality P(yn+1=yi|y)>P(yi|y−i) What does this mean? The finding is intuitively obvious, but poses some challenges to formally defining the notion of overfitting. This is essentially because the ideal amount of fitting for a model to do on some data is nonzero, and overfitting should be "exceeding" this level of ideal fitting. In practice, though, it's difficult to know what is the "appropriate" amount of fitting for a model to be doing. Bayesian inference is ideal if the true model is within the class of models under consideration, but it might fail in unexpected ways if it's not, which is almost always the case in practice. I think the lesson to draw from this is that overfitting is a relative concept and claiming that a particular method "overfits" the data doesn't make too much sense without a point of reference in mind. If people have alternative ways of trying to construct an absolute notion of overfitting with the a...

speech ea guaranteed plugging bayesian bayes interpreted rationalist 2p lesswrong andrew gelman bayesian methods

#78 Exploring MCMC Sampler Algorithms, with Matt D. Hoffman

Play Episode Listen Later Mar 1, 2023 62:41 Transcription Available

Proudly sponsored by PyMC Labs, the Bayesian Consultancy. Book a call, or get in touch!Matt Hoffman has already worked on many topics in his life – music information retrieval, speech enhancement, user behavior modeling, social network analysis, astronomy, you name it.Obviously, picking questions for him was hard, so we ended up talking more or less freely — which is one of my favorite types of episodes, to be honest.You'll hear about the circumstances Matt would advise picking up Bayesian stats, generalized HMC, blocked samplers, why do the samplers he works on have food-based names, etc.In case you don't know him, Matt is a research scientist at Google. Before that, he did a postdoc in the Columbia Stats department, working with Andrew Gelman, and a Ph.D at Princeton, working with David Blei and Perry Cook.Matt is probably best known for his work in approximate Bayesian inference algorithms, such as stochastic variational inference and the no-U-turn sampler, but he's also worked on a wide range of applications, and contributed to software such as Stan and TensorFlow Probability.Our theme music is « Good Bayesian », by Baba Brinkman (feat MC Lars and Mega Ran). Check out his awesome work at https://bababrinkman.com/ !Thank you to my Patrons for making this episode possible!Yusuke Saito, Avi Bryant, Ero Carrera, Giuliano Cruz, Tim Gasser, James Wade, Tradd Salvo, William Benton, James Ahloy, Robin Taylor, Thomas Wiecki, Chad Scherrer, Nathaniel Neitzke, Zwelithini Tunyiswa, Bertrand Wilden, James Thompson, Stephen Oates, Gian Luca Di Tanna, Jack Wells, Matthew Maldonado, Ian Costley, Ally Salim, Larry Gill, Joshua Duncan, Ian Moran, Paul Oreto, Colin Caprani, Colin Carroll, Nathaniel Burbank, Michael Osthege, Rémi Louf, Clive Edelsten, Henri Wallen, Hugo Botha, Vinh Nguyen, Raul Maldonado, Marcin Elantkowski, Adam C. Smith, Will Kurt, Andrew Moskowitz, Hector Munoz, Marco Gorelli, Simon Kessell, Bradley Rode, Patrick Kelley, Rick Anderson, Casper de Bruin, Philippe Labonde, Michael Hankin, Cameron Smith, Tomáš Frýda, Ryan Wesslen, Andreas Netti, Riley King, Yoshiyuki Hamajima, Sven De Maeyer, Michael DeCrescenzo, Fergal M, Mason Yahr, Naoya Kanai, Steven Rowland, Aubrey Clayton, Jeannine Sue, Omri Har Shemesh, Scott Anthony Robson, David Haas, Robert Yolken, Or Duek, Pavel Dusek, Paul Cox, Trey Causey, Andreas Kröpelin, Raphaël R, Nicolas Rode and Gabriel Stechschulte.Visit https://www.patreon.com/learnbayesstats to unlock exclusive Bayesian swag ;)Links from the show:Matt's website: http://matthewdhoffman.com/Matt on Google Scholar: https://scholar.google.com/citations?hl=en&user=IeHKeGYAAAAJ&view_op=list_worksThe No-U-Turn Sampler: Adaptively Setting Path Lengths in Hamiltonian Monte Carlo: https://www.jmlr.org/papers/volume15/hoffman14a/hoffman14a.pdfTuning-Free Generalized Hamiltonian Monte Carlo:

google algorithms hoffman casper rapha proudly sampler bruin bayesian google scholar cameron smith mega ran matt d james thompson hmc james wade paul cox matt hoffman mc lars mcmc baba brinkman rick anderson robin taylor andrew gelman david haas patrick kelley andreas kr aubrey clayton trey causey adam c smith

#76 The Past, Present & Future of Stan, with Bob Carpenter

dm sick statistics arguing objective probability statistical bayesian effective altruism understandable vaden vaclav quanta magazine tetlock andrew gelman bayesianism

Play Episode Listen Later Feb 1, 2023 71:10 Transcription Available

Proudly sponsored by PyMC Labs, the Bayesian Consultancy. Book a call, or get in touch!How does it feel to switch careers and start a postdoc at age 47? How was it to be one of the people who created the probabilistic programming language Stan? What should the Bayesian community focus on in the coming years?These are just a few of the questions I had for my illustrious guest in this episode — Bob Carpenter. Bob is, of course, a Stan developer, and comes from a math background, with an emphasis on logic and computer science theory. He then did his PhD in cognitive and computer sciences, at the University of Edinburgh.He moved from a professor position at Carnegie Mellon to industry research at Bell Labs, to working with Andrew Gelman and Matt Hoffman at Columbia University. Since 2020, he's been working at Flatiron Institute, a non-profit focused on algorithms and software for science.In his free time, Bob loves to cook, see live music, and play role playing games — think Monster of the Week, Blades in Dark, and Fate.Our theme music is « Good Bayesian », by Baba Brinkman (feat MC Lars and Mega Ran). Check out his awesome work at https://bababrinkman.com/ !Thank you to my Patrons for making this episode possible!Yusuke Saito, Avi Bryant, Ero Carrera, Giuliano Cruz, Tim Gasser, James Wade, Tradd Salvo, William Benton, James Ahloy, Robin Taylor, Thomas Wiecki, Chad Scherrer, Nathaniel Neitzke, Zwelithini Tunyiswa, Elea McDonnell Feit, Bert≈rand Wilden, James Thompson, Stephen Oates, Gian Luca Di Tanna, Jack Wells, Matthew Maldonado, Ian Costley, Ally Salim, Larry Gill, Joshua Duncan, Ian Moran, Paul Oreto, Colin Caprani, Colin Carroll, Nathaniel Burbank, Michael Osthege, Rémi Louf, Clive Edelsten, Henri Wallen, Hugo Botha, Vinh Nguyen, Raul Maldonado, Marcin Elantkowski, Adam C. Smith, Will Kurt, Andrew Moskowitz, Hector Munoz, Marco Gorelli, Simon Kessell, Bradley Rode, Patrick Kelley, Rick Anderson, Casper de Bruin, Philippe Labonde, Michael Hankin, Cameron Smith, Tomáš Frýda, Ryan Wesslen, Andreas Netti, Riley King, Yoshiyuki Hamajima, Sven De Maeyer, Michael DeCrescenzo, Fergal M, Mason Yahr, Naoya Kanai, Steven Rowland, Aubrey Clayton, Jeannine Sue, Omri Har Shemesh, Scott Anthony Robson, David Haas, Robert Yolken, Or Duek, Pavel Dusek, Paul Cox, Trey Causey, Andreas Kröpelin and Raphaël R.Visit https://www.patreon.com/learnbayesstats to unlock exclusive Bayesian swag ;)Links from the show:Bob's website: https://bob-carpenter.github.ioBob on GitHub: https://github.com/bob-carpenterBob on Google Scholar: https://scholar.google.com.au/citations?user=kPtKWAwAAAAJ&hl=enStat modeling blog: https://statmodeling.stat.columbia.eduStan home page:

#46 (Bonus) - Arguing about probability (with Nick Anyos)

Increments

Play Episode Listen Later Dec 19, 2022 119:16

We make a guest appearance on Nick Anyos' podcast to talk about effective altruism, longtermism, and probability. Nick (very politely) pushes back on our anti-Bayesian credo, and we get deep into the weeds of probability and epistemology. You can find Nick's podcast on institutional design here (https://institutionaldesign.podbean.com/), and his substack here (https://institutionaldesign.substack.com/?utm_source=substack&utm_medium=web&utm_campaign=substack_profile). We discuss: - The lack of feedback loops in longtermism - Whether quantifying your beliefs is helpful - Objective versus subjective knowledge - The difference between prediction and explanation - The difference between Bayesian epistemology and Bayesian statistics - Statistical modelling and when statistics is useful Links - Philosophy and the practice of Bayesian statistics (http://www.stat.columbia.edu/~gelman/research/published/philosophy.pdf) by Andrew Gelman and Cosma Shalizi - EA forum post (https://forum.effectivealtruism.org/posts/hqkyaHLQhzuREcXSX/data-on-forecasting-accuracy-across-different-time-horizons#Calibrations) showing all forecasts beyond a year out are uncalibrated. - Vaclav smil quote where he predicts a pandemic by 2021: > The following realities indicate the imminence of the risk. The typical frequency of influenza pan- demics was once every 50–60 years between 1700 and 1889 (the longest known gap was 52 years, between the pandemics of 1729–1733 and 1781–1782) and only once every 10–40 years since 1889. The recurrence interval, calculated simply as the mean time elapsed between the last six known pandemics, is about 28 years, with the extremes of 6 and 53 years. Adding the mean and the highest interval to 1968 gives a span between 1996 and 2021. We are, probabilistically speaking, very much inside a high-risk zone. > > - Global Catastropes and Trends, p.46 Reference for Tetlock's superforecasters failing to predict the pandemic. "On February 20th, Tetlock's superforecasters predicted only a 3% chance that there would be 200,000+ coronavirus cases a month later (there were)." (https://wearenotsaved.com/2020/04/18/pandemic-uncovers-the-ridiculousness-of-superforecasting/) Contact us - Follow us on Twitter at @IncrementsPod, @BennyChugg, @VadenMasrani - Check us out on youtube at https://www.youtube.com/channel/UC_4wZzQyoW4s4ZuE4FY9DQQ - Come join our discord server! DM us on twitter or send us an email to get a supersecret link Errata - At the beginning of the episode Vaden says he hasn't been interviewed on another podcast before. He forgot his appearence (https://www.thedeclarationonline.com/podcast/2019/7/23/chesto-and-vaden-debatecast) on The Declaration Podcast in 2019, which will be appearing as a bonus episode on our feed in the coming weeks. Sick of hearing us talk about this subject? Understandable! Send topic suggestions over to incrementspodcast@gmail.com. Photo credit: James O'Brien (http://www.obrien-studio.com/) for Quanta Magazine (https://www.quantamagazine.org/where-quantum-probability-comes-from-20190909/)

Andrew Gelman on Voting, Income, and the Red-state, Blue-state Paradox

EconoFact Chats

Play Episode Listen Later Nov 6, 2022 22:01

There's evidence that economically better-off voters tilt Republican. But there is a paradox. While richer voters tilt Republican, richer states tend to vote Democrat. To discuss this apparent paradox, as well as issues of poll accuracy, and how much the state of the economy has mattered in recent mid-term elections, EconoFact Chats welcomes Andrew Gelman of Columbia University. Andrew is a professor of statistics and political science at Columbia. His work has focused on a range of topics, including why it is rational to vote, and why campaign polls are so variable, when elections are often predictable.

republicans democrats columbia voting income columbia university paradoxes blue states red state blue state andrew gelman

Andrew Gelman on Voting, Income, and the Red-state, Blue-state Paradox

EconoFact Chats

Play Episode Listen Later Nov 6, 2022 22:01

republicans democrats columbia voting income columbia university paradoxes blue states red state blue state andrew gelman

167 | Visualization and Statistics with Andrew Gelman and Jessica Hullman

Data Stories

Play Episode Listen Later Oct 6, 2022 49:26 Very Popular

167 | Visualization and Statistics with Andrew Gelman and Jessica Hullman

statistics visualizations andrew gelman

Bayesian thinking in work and life, ad attribution models and A/B testing, machine learning@Foursquare - Max Sklar - the data scientist show050

The Data Scientist Show

Play Episode Listen Later Sep 13, 2022 90:25

Max Sklar is an independent engineer and researcher. Previously, he was an engineering and Innovation Labs Advisor at Foursquare after 7 years at the company as a machine learning engineer. Previously, he has worked on Ad Attribution, recommendation engine, ratings. He is the host of The Local Maximum podcast. Max studied CS from Yale, and holds a Master degree in information systems from New York university. If you like the show, subscribe to the channel and give us a 5-star review :) Follow Daliana on https://twitter.com/DalianaLiu for more on data science and this podcast. Max's Linkedin: https://www.linkedin.com/in/max-sklar-b638464/ Max's website: localmaxradio.com/about Interviews he mentioned during the podcast: Andrew Gelman, Statistics at Columbia University Shirin Mojarad on Causality Johnny Nelson on Free Speech and Moderation online Stephanie Yang talking about Foursquare's Venue Rating System Dennis Crowley: on Labs, on Innovation Sophie Carr (Bayesian Mathematician) Will Kurt (Bayesian) Marsbot for Airpods Other Episodes Mentioned Bayesian Thinking P-Hacking Interview on Learn Bayesian Statistics

new york master thinking testing yale models statistics machine learning free speech labs cs data scientists attribution foursquare work and life bayesian sklar andrew gelman local maximum

Andrew Gelman, PhD - Election Forecasting

Dana-Farber Data Science Podcast

Play Episode Listen Later May 11, 2022 47:56

How is statistics used to predict elections? Andrew and Rafa discuss the U.S. 2020 Election and the role of the electoral college, polls, mail-in ballots and voter data in forecasting results and post-election outcomes. Andrew Gelman, PhD is a professor of statistics and political science at Columbia University. He is one of the go-to statisticians for the New York Times and author of perhaps the most popular statistics blog: Statistical Modeling, Causal Inference, and Social Science. He has received the Outstanding Statistical Application award three times from the American Statistical Association, the award for best article published in the American Political Science Review, and the Council of Presidents of Statistical Societies award for outstanding contributions by a person under the age of 40. Books he has authored and co-authored include Bayesian Data Analysis, Teaching Statistics: A Bag of Tricks, Data Analysis Using Regression and Multilevel/Hierarchical Models, Red State, Blue State, Rich State, Poor State: Why Americans Vote the Way They Do, A Quantitative Tour of the Social Sciences, and Regression and Other Stories. Our Data Science Zoominars feature interactive conversation with data science experts and a Q+A session moderated by Rafael A. Irizarry, PhD, Chair, Department of Data Science at Dana-Farber Cancer Institute.

EA - The accidental experiment that saved 700 lives (IRS & health insurance) by Lizka

The Nonlinear Library

Play Episode Listen Later Apr 19, 2022 3:28

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: The accidental experiment that saved 700 lives (IRS & health insurance), published by Lizka on April 19, 2022 on The Effective Altruism Forum. From Lizka: I really enjoy the blog, "Statistical Modeling, Causal Inference, and Social Science." Andrew Gelman, one of the authors of the blog, has given me permission to cross-post this post, which I thought some Forum readers might find interesting. As an aside, I like many other posts on the blog. Two examples are "Parables vs. stories" and "The social sciences are useless. So why do we study them? Here's a good reason:." Paul Alper sends along this news article by Sarah Kliff, who writes: Three years ago, 3.9 million Americans received a plain-looking envelope from the Internal Revenue Service. Inside was a letter stating that they had recently paid a fine for not carrying health insurance and suggesting possible ways to enroll in coverage. . . . Three Treasury Department economists [Jacob Goldin, Ithai Lurie, and Janet McCubbin] have published a working paper finding that these notices increased health insurance sign-ups. Obtaining insurance, they say, reduced premature deaths by an amount that exceeded any of their expectations. Americans between 45 and 64 benefited the most: For every 1,648 who received a letter, one fewer death occurred than among those who hadn't received a letter. . . . The experiment, made possible by an accident of budgeting, is the first rigorous experiment to find that health coverage leads to fewer deaths, a claim that politicians and economists have fiercely debated in recent years as they assess the effects of the Affordable Care Act's coverage expansion. The results also provide belated vindication for the much-despised individual mandate that was part of Obamacare until December 2017, when Congress did away with the fine for people who don't carry health insurance. “There has been a lot of skepticism, especially in economics, that health insurance has a mortality impact,” said Sarah Miller, an assistant professor at the University of Michigan who researches the topic and was not involved with the Treasury research. “It's really important that this is a randomized controlled trial. It's a really high standard of evidence that you can't just dismiss.” This graph shows how the treatment increased health care coverage during the months after it was applied: And here's the estimated effect on mortality: They should really label the lines directly. Sometimes it seems that economists think that making a graph easier to read is a form of cheating! I'd also like to see some multilevel modeling—as it is, they end up with lots of noisy estimates, lots of wide confidence intervals, and I think more could be done. But that's fine. It's best that the authors did what they did, which was to present their results. Now that the data are out there, other researchers can go back in and do more sophisticated analysis. That's how research should go. It would not make sense for such important results to be held under wraps, waiting for some ideal statistical analysis that might never happens. Overall, this is an inspiring story of what can be learned from a natural experiment. The news article also has this sad conclusion: At the end of 2017, Congress passed legislation eliminating the health law's fines for not carrying health insurance, a change that probably guarantees that the I.R.S. letters will remain a one-time experiment. But now that they have evidence that the letters had a positive effect, maybe they'll restart the program, no? Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.

Chris Tosh | The piranha problem in statistics

statistics declaration confusing bounds piranhas tosh bayesian andrew gelman

Play Episode Listen Later Feb 22, 2022 69:41 Very Popular

The piranha problem (too many large, independent effect sizes influence the same outcome) has received some attention on Andrew Gelman's blog. But now it's a paper! Chris Tosh (Memorial Sloan Kettering) talks about multiple views of the piranha problem and detecting the implausible scientific claims that are published. The butterfly effect makes an appearance. If you enjoyed the science-vs-pseudoscience topics, you'll enjoy this one. 0:00 - Coming up in the episode 2:35 - What is the Piranha Problem? 19:54 - Confusing effect sizes 23:11 - The "words & walking speed" study 26:22 - Declaration of independent variables 30:58 - Piranha theorems for correlations 37:07 - Piranha theorems for linear regression 40:37 - Piranha Theorems for mutual information 44:13 - Bounds on the independence of the covariates 46:12 - Applying the piranha theorem to real data 50:12 - Applying the piranha theorem across studies 54:05 - A Bayesian detour 1:00:12 - The butterfly effect & chaos 1:04:26 - Applying the piranha theorem to cancer research

Some personal thoughts on EA and systemic change by CarlShulman

The Nonlinear Library: EA Forum Top Posts

Play Episode Listen Later Dec 12, 2021 10:32

welcome to the nonlinear library, where we use text-to-speech software to convert the best writing from the rationalist and ea communities into audio. this is: Some personal thoughts on EA and systemic change, published by CarlShulman on the effective altruism forum. DavidNash requested that I repost my comment below, on what to make of discussions about EA neglecting systemic change, as a top-level post. These are my off-the-cuff thoughts and no one else's. In summary (to be unpacked below): Actual EA is able to do assessments of systemic change interventions including electoral politics and policy change, and has done so a number of times The great majority of critics of EA invoking systemic change fail to present the simple sort of quantitative analysis given above for the interventions they claim excel, and frequently when such analysis is done the intervention does not look competitive by EA lights Nonetheless, my view is that historical data do show that the most efficient political/advocacy spending, particularly aiming at candidates and issues selected with an eye to global poverty or the long term, does have higher returns than GiveWell top charities (even ignoring nonhumans and future generations or future technologies); one can connect the systemic change critique as a position in intramural debates among EAs about the degree to which one should focus on highly linear, giving as consumption, type interventions EAs who are willing to consider riskier and less linear interventions are mostly already pursuing fairly dramatic systemic change, in areas with budgets that are small relative to political spending (unlike foreign aid) As funding expands in focused EA priority issues, eventually diminishing returns there will equalize with returns for broader political spending, and activity in the latter area could increase enormously: since broad political impact per dollar is flatter over a large range political spending should either be a very small or very large portion of EA activity In full: Actual EA is able to do assessments of systemic change interventions including electoral politics and policy change, and has done so a number of times Empirical data on the impact of votes, the effectiveness of lobbying and campaign spending work out without any problems of fancy decision theory or increasing marginal returns E.g. Andrew Gelman's data on US Presidential elections shows that given polling and forecasting uncertainty a marginal vote in a swing state average something like a 1 in 10 million chance of swinging an election over multiple elections (and one can save to make campaign contributions 80,000 Hours has a page (there have been a number of other such posts and discussion, note that 'worth voting' and 'worth buying a vote through campaign spending or GOTV' are two quite different thresholds) discussing this data and approaches to valuing differences in political outcomes between candidates; these suggest that a swing state vote might be worth tens of thousands of dollars of income to rich country citizens But if one thinks that charities like AMF do 100x or more good per dollar by saving the lives of the global poor so cheaply, then these are compatible with a vote being worth only a few hundred dollars If one thinks that some other interventions, such as gene drives for malaria eradication, animal advocacy, or existential risk interventions are much more cost-effective than AMF, that would lower the value further except insofar as one could identify strong variation in more highly-valued effects Experimental data on the effects of campaign contributions suggest a cost of a few hundred dollars per marginal vote (see, e.g. Gerber's work on GOTV experiments) Prediction markets and polling models give a good basis for assessing the chance of billions of dollars of campaign funds swinging an election If there are increasing returns to scale from large-scale spending, small donors can convert their funds into a smal...

personal predictions ea gerber us presidential eas systemic change amf gotv givewell personal thoughts andrew gelman

#52 Election forecasting models in Germany, with Marcus Gross

research columbia statistical causal descriptive andrew gelman

Play Episode Listen Later Dec 9, 2021 58:08

Did I mention I like survey data, especially in the context of electoral forecasting? Probably not, as I'm a pretty shy and reserved man. Why are you laughing?? Yeah, that's true, I'm not that shy… but I did mention my interest for electoral forecasting already! And before doing a full episode where I'll talk about French elections (yes, that'll come at one point), let's talk about one of France's neighbors — Germany. Our German friends had federal elections a few weeks ago — consequential elections, since they had the hard task of replacing Angela Merkel, after 16 years in power. To talk about this election, I invited Marcus Gross on the show, because he worked on a Bayesian forecasting model to try and predict the results of this election — who will get elected as Chancellor, by how much and with which coalition? I was delighted to ask him about how the model works, how it accounts for the different sources of uncertainty — be it polling errors, unexpected turnout or media events — and, of course, how long it takes to sample (I think you'll be surprised by the answer). We also talked about the other challenge of this kind of work: communication — how do you communicate uncertainty effectively? How do you differentiate motivated reasoning from useful feedback? What were the most common misconceptions about the model? Marcus studied statistics in Munich and Berlin, and did a PhD on survey statistics and measurement error models in economics and archeology. He worked as a data scientist at INWT, a consulting firm with projects in different business fields as well as the public sector. Now, he is working at FlixMobility. Our theme music is « Good Bayesian », by Baba Brinkman (feat MC Lars and Mega Ran). Check out his awesome work at https://bababrinkman.com/ (https://bababrinkman.com/) ! Thank you to my Patrons for making this episode possible! Yusuke Saito, Avi Bryant, Ero Carrera, Brian Huey, Giuliano Cruz, Tim Gasser, James Wade, Tradd Salvo, Adam Bartonicek, William Benton, Alan O'Donnell, Mark Ormsby, Demetri Pananos, James Ahloy, Robin Taylor, Thomas Wiecki, Chad Scherrer, Nathaniel Neitzke, Zwelithini Tunyiswa, Elea McDonnell Feit, Bertrand Wilden, James Thompson, Stephen Oates, Gian Luca Di Tanna, Jack Wells, Matthew Maldonado, Ian Costley, Ally Salim, Larry Gill, Joshua Duncan, Ian Moran, Paul Oreto, Colin Caprani, George Ho, Colin Carroll, Nathaniel Burbank, Michael Osthege, Rémi Louf, Clive Edelsten, Henri Wallen, Hugo Botha, Vinh Nguyen, Raul Maldonado, Marcin Elantkowski, Adam C. Smith, Will Kurt, Andrew Moskowitz, Hector Munoz, Marco Gorelli, Simon Kessell, Bradley Rode, Patrick Kelley, Rick Anderson, Casper de Bruin, Philippe Labonde, Matthew McAnear, Michael Hankin, Cameron Smith, Luis Iberico, Alejandro Morales, Tomáš Frýda, Ryan Wesslen, Andreas Netti, Riley King and Aaron Jones. Visit https://www.patreon.com/learnbayesstats (https://www.patreon.com/learnbayesstats) to unlock exclusive Bayesian swag ;) Links from the show: German election forecast website: https://www.wer-gewinnt-die-wahl.de/en (https://www.wer-gewinnt-die-wahl.de/en) Twitter account of electoral model: https://twitter.com/GerElectionFcst (https://twitter.com/GerElectionFcst) German election model code: https://github.com/INWTlab/lsTerm-election-forecast (https://github.com/INWTlab/lsTerm-election-forecast) LBS #27 -- Modeling the US Presidential Elections, with Andrew Gelman & Merlin Heidemanns: https://www.learnbayesstats.com/episode/27-modeling-the-us-presidential-elections-with-andrew-gelman-merlin-heidemanns (https://www.learnbayesstats.com/episode/27-modeling-the-us-presidential-elections-with-andrew-gelman-merlin-heidemanns) This podcast uses the following third-party services for analysis: Podcorn - https://podcorn.com/privacy Support this podcast

COVID, ALIENS & THE ARCHONS - Robert Bonomo X Michael Decon

End of Days

Play Episode Listen Later Dec 5, 2021 78:19

Episode 355: The Archons In the last decade, we have seen an exponential change in the manipulation of basic human instincts through a technological and societal shift often referred to as the Fourth Industrial Revolution. Everything we do, from sex, dating, and purchases to political affiliations and how we define ourselves is being manipulated and commoditized. Money is the air of civilization, necessary for just about everything we do outside breathing, but many people still lack a basic understanding of what it is and how it is created. How can we have a democracy if our most basic of necessities is almost universally misunderstood? One would have thought, before 2020, that virology and epidemiology were inherently apolitical. Incredibly, even one's views on antibody tests are highly correlated with voting preference and whether one prefers Tucker Carlson to Rachel Maddow. By the time the now-infamous Stanford-led Santa Clara antibody tests came out, I could already predict how my blue checkmark list of Twitter experts was going to respond. 'I think the authors owe us all an apology… not just to us, but to Stanford,' wrote Andrew Gelman, a professor of statistics and political science and director of the Applied Statistics Center at Columbia University. Dr. Gelman is a serious clock, and he was furious. How could anyone from a reputable university even begin to suggest that we weren't facing the greatest threat to humanity since the Black Death? The first wave of antibody test results implied that the infection fatality rate (IFR) was lower than first reported and more carriers were asymptomatic as well. This was NOT good news for the clocks; it was pure heresy.

covid-19 money aliens stanford columbia university tucker carlson incredibly santa clara black death rachel maddow fourth industrial revolution archons ifr gelman andrew gelman robert bonomo michael decon

Ep. 201 - Andrew Gelman, Statistical Insights

The Local Maximum

Play Episode Listen Later Nov 30, 2021 33:51

Max interviews Professor Andrew Gelman of Columbia on Bayesian Inference, Research, Political Prognostication, Descriptive vs Causal work, and areas of computational research. localmaxradio.com/201

#45 Biostats & Clinical Trial Design, with Frank Harrell

politics professor vote columbia university statistics american politics predictable variable andrew gelman

Play Episode Listen Later Aug 10, 2021 68:54

As a podcaster, I discovered that there are guests for which the hardest is to know when to stop the conversation. They could talk for hours and that would make for at least 10 fantastic episodes. Frank Harrell is one of those guests. To me, our conversation was both fascinating — thanks to Frank's expertise and the width and depth of topics we touched on — and frustrating — I still had a gazillion questions for him! But rest assured, we talked about intent to treat and randomization, proportional odds, clinical trial design, bio stats and covid19, and even which mistakes you should do to learn Bayes stats — yes, you heard right, which mistakes. Anyway, I can't tell you everything here — you'll just have to listen to the episode! A long time Bayesian, Frank is a Professor of Biostatistics in the School of Medicine at Vanderbilt University. His numerous research interests include predictive models and model validation, Bayesian clinical trial design and Bayesian models, drug development, and clinical research. He holds a PhD in biostatistics from the University of North Carolina, and did his Bachelor in mathematics at the University of Alabama in Birmingham. Our theme music is « Good Bayesian », by Baba Brinkman (feat MC Lars and Mega Ran). Check out his awesome work at https://bababrinkman.com/ (https://bababrinkman.com/) ! Thank you to my Patrons for making this episode possible! Yusuke Saito, Avi Bryant, Ero Carrera, Brian Huey, Giuliano Cruz, Tim Gasser, James Wade, Tradd Salvo, Adam Bartonicek, William Benton, Alan O'Donnell, Mark Ormsby, Demetri Pananos, James Ahloy, Jon Berezowski, Robin Taylor, Thomas Wiecki, Chad Scherrer, Nathaniel Neitzke, Zwelithini Tunyiswa, Elea McDonnell Feit, Bertrand Wilden, James Thompson, Stephen Oates, Gian Luca Di Tanna, Jack Wells, Matthew Maldonado, Ian Costley, Ally Salim, Larry Gill, Joshua Duncan, Ian Moran, Paul Oreto, Colin Caprani, George Ho, Colin Carroll, Nathaniel Burbank, Michael Osthege, Rémi Louf, Clive Edelsten, Henri Wallen, Hugo Botha, Vinh Nguyen, Raul Maldonado, Marcin Elantkowski, Tim Radtke, Adam C. Smith, Will Kurt, Andrew Moskowitz, Hector Munoz, Marco Gorelli, Simon Kessell, Bradley Rode, Patrick Kelley, Rick Anderson, Casper de Bruin and Philippe Labonde. Visit https://www.patreon.com/learnbayesstats (https://www.patreon.com/learnbayesstats) to unlock exclusive Bayesian swag ;) Links from the show: Frank's website and courses: https://hbiostat.org/ (https://hbiostat.org/) Frank's blog: https://www.fharrell.com/ (https://www.fharrell.com/) Frank on Twitter: https://twitter.com/f2harrell (https://twitter.com/f2harrell) COVID-19 Randomized Clinical Trial Design: https://hbiostat.org/proj/covid19/ (https://hbiostat.org/proj/covid19/) Frank on GitHub: https://github.com/harrelfe (https://github.com/harrelfe) Regression Modeling Strategies repository: https://github.com/harrelfe/rms (https://github.com/harrelfe/rms) Biostatistics for Biomedical Research repository: https://github.com/harrelfe/bbr (https://github.com/harrelfe/bbr) Bayesian Approaches to Randomized Trials, Spiegelhalter et al.: http://hbiostat.org/papers/Bayes/spi94bay.pdf (http://hbiostat.org/papers/Bayes/spi94bay.pdf) Statistical Rethinking, Richard McElreath: http://xcelab.net/rm/statistical-rethinking/ (http://xcelab.net/rm/statistical-rethinking/) LBS #20, Regression and Other Stories, with Andrew Gelman, Jennifer Hill & Aki Vehtari: https://www.learnbayesstats.com/episode/20-regression-and-other-stories-with-andrew-gelman-jennifer-hill-aki-vehtari (https://www.learnbayesstats.com/episode/20-regression-and-other-stories-with-andrew-gelman-jennifer-hill-aki-vehtari) David Spiegelhalter, The Art of Statistics -- Learning from Data: https://www.amazon.fr/Art-Statistics-Learning-Data/dp/0241398630 (https://www.amazon.fr/Art-Statistics-Learning-Data/dp/0241398630) This podcast uses the following third-party services for analysis: Podcorn -... Support this podcast

Statistics and American Politics with Andrew Gelman of Columbia University

The Great Battlefield

Play Episode Listen Later Jun 14, 2021 60:38

Professor of Statistics and Politics, Andrew Gelman joins The Great Battlefield podcast to talk about his career including a political science paper he co-wrote called "Why are American Presidential Campaign Polls so Variable when Votes are so Predictable".

How Sean Taylor of Lyft Rideshare Labs thinks about business decision problems

Gradient Dissent - A Machine Learning Podcast by W&B

Play Episode Listen Later May 13, 2021 45:40

Sean joins us to chat about ML models and tools at Lyft Rideshare Labs, Python vs R, time series forecasting with Prophet, and election forecasting. --- Sean Taylor is a Data Scientist at (and former Head of) Lyft Rideshare Labs, and specializes in methods for solving causal inference and business decision problems. Previously, he was a Research Scientist on Facebook's Core Data Science team. His interests include experiments, causal inference, statistics, machine learning, and economics. Connect with Sean: Personal website: https://seanjtaylor.com/ Twitter: https://twitter.com/seanjtaylor LinkedIn: https://www.linkedin.com/in/seanjtaylor/ --- Topics Discussed: 0:00 Sneak peek, intro 0:50 Pricing algorithms at Lyft 07:46 Loss functions and ETAs at Lyft 12:59 Models and tools at Lyft 20:46 Python vs R 25:30 Forecasting time series data with Prophet 33:06 Election forecasting and prediction markets 40:55 Comparing and evaluating models 43:22 Bottlenecks in going from research to production Transcript: http://wandb.me/gd-sean-taylor Links Discussed: "How Lyft predicts a rider’s destination for better in-app experience"": https://eng.lyft.com/how-lyft-predicts-your-destination-with-attention-791146b0a439 Prophet: https://facebook.github.io/prophet/ Andrew Gelman's blog post "Facebook's Prophet uses Stan": https://statmodeling.stat.columbia.edu/2017/03/01/facebooks-prophet-uses-stan/ Twitter thread "Election forecasting using prediction markets": https://twitter.com/seanjtaylor/status/1270899371706466304 "An Updated Dynamic Bayesian Forecasting Model for the 2020 Election": https://hdsr.mitpress.mit.edu/pub/nw1dzd02/release/1 --- Get our podcast on these platforms: Apple Podcasts: http://wandb.me/apple-podcasts Spotify: http://wandb.me/spotify Google Podcasts: http://wandb.me/google-podcasts YouTube: http://wandb.me/youtube Soundcloud: http://wandb.me/soundcloud Join our community of ML practitioners where we host AMAs, share interesting projects and meet other people working in Deep Learning: http://wandb.me/slack Check out Fully Connected, which features curated machine learning reports by researchers exploring deep learning techniques, Kagglers showcasing winning models, industry leaders sharing best practices, and more: https://wandb.ai/fully-connected

Andrew Gelman & Megan Higgs | Statistics' Role in Science and Pseudoscience

science math statistics higgs pseudoscience fooling andrew gelman

Play Episode Listen Later Apr 30, 2021 71:53 Very Popular

Andrew Gelman & Megan Higgs | Statistics' Role in Science and Pseudoscience #datascience #statistics #science #pseudoscience Our science vs pseudoscience discussion continues with Andrew Gelman (Columbia) and Megan Higgs (Critical Inference LLC). Andrew and Megan describe two critical roles that statistics plays in science.... but also how statistics can add the air of scientific rigor to bad research or help statisticians fool themselves. From there the conversation goes on in a way that only a conversation with Andrew and Megan can! A very fun episode. 0:00 - Two roles of statistics in science 4:50 - Many models were intended for designed experiments 10:30 - The biggest scientific error of the past 20 years 15:00 - Feedback loop of over-confidence / Armstrong Principle 21:00 - Science is personal 25:00 - The value of different approaches / Don Rubin Story 34:40 - Statistics is the science of defaults / engineering new methods 45:00 - The value of writing what you did 52:27 - Math vs science backgrounds + a thought experiment 1:01:20 - Fooling ourselves

Stalking the billion-footed beast II (4-18-21)

Luke Ford

Play Episode Listen Later Apr 18, 2021 135:39

00:00 Dr. David Starkey: I Was Cancelled but I Won't be Silenced for Speaking Objective Truth, https://www.youtube.com/watch?v=UrDOkYGd5d8 03:00 Tom Wolfe, Stalking the Billion Footed Beast, https://harpers.org/archive/1989/11/stalking-the-billion-footed-beast/ 14:00 Niall Ferguson and the perils of playing to your audience, https://statmodeling.stat.columbia.edu/2018/12/05/niall-ferguson-perils-playing-audience/ 18:00 Pick a title for Niall Ferguson's next book!, https://statmodeling.stat.columbia.edu/2015/10/11/new-competition-pick-a-title-for-niall-fergusons-next-book/ 21:00 Going meta on Niall Ferguson, https://statmodeling.stat.columbia.edu/2013/07/01/going-meta-on-niall-ferguson/ 24:00 Andrew Gelman on Niall Ferguson, https://statmodeling.stat.columbia.edu/?s=%22niall+ferguson%22 29:00 The Real Problem with Niall Ferguson's Letter to the 1%, https://www.esquire.com/news-politics/a31282/niall-ferguson-newsweek-cover-11914269/ 32:00 Niall Ferguson, the John Yoo line, and the paradox of influence, https://statmodeling.stat.columbia.edu/2012/09/12/niall-ferguson-the-john-yoo-line-and-the-paradox-of-influence/ 33:00 The John Yoo line, https://themonkeycage.org/2012/09/niall-ferguson-crosses-the-john-yoo-line-the-paradox-of-influence/ 37:00 Greg Conte and the National Justice Party, https://odysee.com/@modernpolitics:0/ModPol-ContePart1:4 56:00 How Philanthropy Is Fueling American Division, https://www.nationalreview.com/2021/04/how-philanthropy-is-fueling-american-division/ 1:12:00 Suspected FedEx shooter was part of My Little Pony 'brony' subculture, https://thepostmillennial.com/fedex-shooter-was-part-of-my-little-pony-brony-subculture 1:16:00 Interview with Greg Conte: Part Two, https://odysee.com/@modernpolitics:0/ModPol-ContePart2:1 1:21:00 Propaganda, https://en.wikipedia.org/wiki/Propaganda 1:34:00 A Critique of Ron Unz's Article “The Myth of American Meritocracy”, https://sites.google.com/site/nuritbaytch/ 1:36:00 Janet Mertz on Ron Unz's “Meritocracy”, https://statmodeling.stat.columbia.edu/wp-content/uploads/2013/03/Mertz-on-Unz-Meritocracy-Article.pdf 2:00:00 The dirty tricks and shady tactics of Adam Curtis, https://lwlies.com/articles/adam-curtis-hypernormalisation-tricks-and-tactics/ Effective Communication Skills, https://www.audible.com/pd/Effective-Communication-Skills-Audiobook/B00D94332Q Join this channel to get access to perks: https://www.youtube.com/channel/UCSFVD7Xfhn7sJY8LAIQmH8Q/join https://odysee.com/@LukeFordLive, https://lbry.tv/@LukeFord, https://rumble.com/lukeford https://dlive.tv/lukefordlivestreams Listener Call In #: 1-310-997-4596 Superchat: https://entropystream.live/app/lukefordlive Bitchute: https://www.bitchute.com/channel/lukeford/ Soundcloud MP3s: https://soundcloud.com/luke-ford-666431593 Code of Conduct: https://lukeford.net/blog/?p=125692 https://www.patreon.com/lukeford http://lukeford.net Email me: lukeisback@gmail.com or DM me on Twitter.com/lukeford Support the show | https://www.streamlabs.com/lukeford, https://patreon.com/lukeford, https://PayPal.Me/lukeisback Facebook: http://facebook.com/lukecford Feel free to clip my videos. It's nice when you link back to the original.

#34 Multilevel Regression, Post-stratification & Missing Data, with Lauren Kennedy

american original views long term cold war statistics political science wide hawkins tipping points polarization research director logically gelman stephen hawkins andrew gelman

Play Episode Listen Later Feb 25, 2021 72:39

Episode sponsored by Tidelift: https://tidelift.com/ (tidelift.com) We already mentioned multilevel regression and post-stratification (MRP, or Mister P) on this podcast, but we didn’t dedicate a full episode to explaining how it works, why it’s useful to deal with non-representative data, and what its limits are. Well, let’s do that now, shall we? To that end, I had the delight to talk with Lauren Kennedy! Lauren is a lecturer in Business Analytics at Monash University in Melbourne, Australia, where she develops new statistical methods to analyze social science data. Working mainly with R and Stan, Lauren studies non-representative data, multilevel modeling, post-stratification, causal inference, and, more generally, how to make inferences from the social sciences. Needless to say that I asked her everything I could about MRP, including how to choose priors, why her recent paper about structured priors can improve MRP, and when MRP is not useful. We also talked about missing data imputation, and how all these methods relate to causal inference in the social sciences. If you want a bit of background, Lauren did her Undergraduates in Psychological Sciences and Maths and Computer Sciences at Adelaide University, with Danielle Navarro and Andrew Perfors, and then did her PhD with the same advisors. She spent 3 years in NYC with Andrew Gelman’s Lab at Columbia University, and then moved back to Melbourne in 2020. Most importantly, Lauren is an adept of crochet — she’s already on her third blanket! Our theme music is « Good Bayesian », by Baba Brinkman (feat MC Lars and Mega Ran). Check out his awesome work at https://bababrinkman.com/ (https://bababrinkman.com/) ! Thank you to my Patrons for making this episode possible! Yusuke Saito, Avi Bryant, Ero Carrera, Brian Huey, Giuliano Cruz, Tim Gasser, James Wade, Tradd Salvo, Adam Bartonicek, William Benton, Alan O'Donnell, Mark Ormsby, Demetri Pananos, James Ahloy, Jon Berezowski, Robin Taylor, Thomas Wiecki, Chad Scherrer, Vincent Arel-Bundock, Nathaniel Neitzke, Zwelithini Tunyiswa, Elea McDonnell Feit, Bertrand Wilden, James Thompson, Stephen Oates, Gian Luca Di Tanna, Jack Wells, Matthew Maldonado, Ian Costley, Ally Salim, Larry Gill, Joshua Duncan, Ian Moran, Paul Oreto, Colin Caprani, George Ho, Colin Carroll and Nathaniel Burbank. Visit https://www.patreon.com/learnbayesstats (https://www.patreon.com/learnbayesstats) to unlock exclusive Bayesian swag ;) Links from the show: Lauren's website: https://jazzystats.com/ (https://jazzystats.com/) Lauren on Twitter: https://twitter.com/jazzystats (https://twitter.com/jazzystats) Lauren on GitHub: https://github.com/lauken13 (https://github.com/lauken13) Improving multilevel regression and poststratification with structured priors: https://arxiv.org/abs/1908.06716 (https://arxiv.org/abs/1908.06716) Using model-based regression and poststratification to generalize findings beyond the observed sample: https://arxiv.org/abs/1906.11323 (https://arxiv.org/abs/1906.11323) Lauren's beginners Bayes workshop: https://github.com/lauken13/Beginners_Bayes_Workshop (https://github.com/lauken13/Beginners_Bayes_Workshop) MRP in RStanarm: https://github.com/lauken13/rstanarm/blob/master/vignettes/mrp.Rmd (https://github.com/lauken13/rstanarm/blob/master/vignettes/mrp.Rmd) Choosing your rstanarm prior with prior predictive checks: https://github.com/stan-dev/rstanarm/blob/vignette-prior-predictive/vignettes/prior-pred.Rmd (https://github.com/stan-dev/rstanarm/blob/vignette-prior-predictive/vignettes/prior-pred.Rmd) Mister P -- What’s its secret sauce?: https://statmodeling.stat.columbia.edu/2013/10/09/mister-p-whats-its-secret-sauce/ (https://statmodeling.stat.columbia.edu/2013/10/09/mister-p-whats-its-secret-sauce/) Bayesian Multilevel Estimation with Poststratification -- State-Level Estimates from National Polls: https://pdfs.semanticscholar.org/2008/bee9f8c2d7e41ac9c5c54489f41989a0d7ba.pdf... Support this podcast

328-A Canine Prisoner of War

Futility Closet

Play Episode Listen Later Jan 18, 2021 30:47

In 1944, British captives of the Japanese in Sumatra drew morale from an unlikely source: a purebred English pointer who cheered the men, challenged the guards, and served as a model of patient fortitude. In this week's episode of the Futility Closet podcast we'll tell the story of Judy, the canine POW of World War II. We'll also consider the frequency of different birthdays and puzzle over a little sun. Intro: Sherlock Holmes wrote 20 monographs. In 1863, Charles Dickens' hall clock stopped sounding. Sources for our feature on Judy: Robert Weintraub, No Better Friend: One Man, One Dog, and Their Incredible Story of Courage and Survival in World War II, 2016. S.L. Hoffman, "Judy: The Unforgettable Story of the Dog Who Went to War and Became a True Hero," Military History 32:1 (May 2015), 72-72. Rebecca Frankel, "Dogs at War: Judy, Canine Prisoner of War," National Geographic, May 18, 2014. Robert Weintraub, "The True Story of Judy, the Dog Who Inspired Her Fellow Prisoners of War to Survive," Irish Times, June 2, 2015. Jane Dalton, "Judy, the Life-Saving PoW Who Beat the Japanese," Sunday Telegraph, May 31, 2015. "Heroine Dog's Medal Goes on Display," [Cardiff] Western Mail, Aug. 26, 2006. "Medal Awarded to Dog Prisoner of War Goes on Public Display," Yorkshire Post, Aug. 23, 2006. Amber Turnau, "The Incredible Tale of Frank Williams," Burnaby [B.C.] Now, March 19, 2003. Nicholas Read, "Prison Camp Heroine Judy Was History's Only Bow-Wow PoW," Vancouver Sun, March 12, 2003. "London Salutes Animal Veterans," Charlotte Observer, May 28, 1983. Frank G. Williams, "The Dog That Went to War," Vancouver Sun, April 6, 1974. "Judy, Dog VC, Dies," [Montreal] Gazette, March 23, 1950. "Judy, British War Dog, Dies; to Get Memorial," [Wilmington, Del.] Morning News, March 21, 1950. "The Tale of a V.C. Dog," [Adelaide] Chronicle, Jan. 30, 1947. "Judy to Receive Dogs' V.C.," The Age, May 2, 1946. "Judy: The Dog Who Became a Prisoner of War," gov.uk, July 24, 2015. "Prisoner of War Dog Judy -- PDSA Dickin Medal and Collar to Be Presented to the Imperial War Museum," People's Dispensary for Sick Animals, Aug. 21, 2006. "PDSA Dickin Medal Stories: Judy," PDSA Schools (accessed Jan. 3, 2021). Listener mail: Andrew Gelman et al., "Bayesian Data Analysis (Third Edition)," 1995-2020. "Keynote: Andrew Gelman - Data Science Workflow" (video), Dec. 21, 2017. Becca R. Levy, Pil H. Chung, and Martin D. Slade, "Influence of Valentine's Day and Halloween on Birth Timing," Social Science & Medicine 73:8 (2011), 1246-1248. "Tony Meléndez," Wikipedia (accessed Dec. 24, 2020). "Thalidomide," Wikipedia (accessed Jan. 9, 2020). Neil Vargesson, "Thalidomide-Induced Teratogenesis: History and Mechanisms," Birth Defects Research Part C: Embryo Today: Reviews 105:2 (2015), 140-156. "Biography," tonymelendez.com (accessed Jan. 10, 2021). "Tony Melendez Sings for Pope John Paul II - 1987" (video), Heart of the Nation, Sept. 27, 2016. This week's lateral thinking puzzle was contributed by listener Lucie. Here's a corroborating link (warning -- this spoils the puzzle). You can listen using the player above, download this episode directly, or subscribe on Google Podcasts, on Apple Podcasts, or via the RSS feed at https://futilitycloset.libsyn.com/rss. Please consider becoming a patron of Futility Closet -- you can choose the amount you want to pledge, and we've set up some rewards to help thank you for your support. You can also make a one-time donation on the Support Us page of the Futility Closet website. Many thanks to Doug Ross for the music in this episode. If you have any questions or comments you can reach us at podcast@futilitycloset.com. Thanks for listening!

Polarization at the Tipping Point: The Long Term and Wide-Angled Views

The Purple Principle

Play Episode Listen Later Jan 12, 2021 26:25

A Major Pandemic… Election 2020 that lasted weeks… Election Denial 2020 is still going months later… Insurrection… Impeachment… And possibly more to come. There's been a lot of polarized and polarizing events over the past few months. So it seems time to step back and take a longer range, wider-angle view of partisanship in our DisUnited States. The Purple Principle does that in Episode 20 with featured guests Dr. Andrew Gelman of Columbia University (Departments of Political Science and Statistics) and Stephen Hawkins, Research Director of the international non-profit, More in Common, authors of the seminal study on American political identity, The Hidden Tribes. Dr. Gelman explains how polarization is measured over time with modern statistical techniques, which reveal how seemingly unrelated issue positions can form into partisan constellations. Why, for example, should someone's position on the minimum wage correlate with their view on global warming? Logically, there's little connection. But in our partisan age, these correlations are increasing over time, if not yet fully correlated. Stephen Hawkins of More in Common defines the seven tribal identities identified through extensive psychology-driven polling. More in Common defines the four groups in the American center as “The Exhausted Majority.” Hawkins explains that those suffering from partisan exhaustion tune out from political news while our tribal wings consume more media, thus incentivizing media companies to play to their outrage. What's a polarized nation to do? Hawkins suggests the answer may lie back in the Cold War, when a common enemy solidified American identity. Tune in to learn more about the major trends in polarization and our surprisingly complex political tribalism in Episode 20, “Polarization at the Tipping Point.” Original music composed and created by Ryan Adair Rooney. For show notes and transcript, please visit our website: www.fluentknowledge.com/shows/the-purple-principle/polarization-at-the-tipping-point

2020 Polling Miss and Political Correctness

Nationalism Course podcast

Play Episode Listen Later Jan 5, 2021 121:19

I begin by discussing the 2020 polling miss in the context of political correctness and falling Republican trust in institutions. This is followed by discussion with leading American social scientists Doug Massey, Andrew Gelman and others. From What Happens Next in 6 Minutes podcast https://whathappensnextin6minutes.podbean.com/ Episode: https://www.podbean.com/site/EpisodeDownload/PBF4EBBEHSNGC

american republicans polling political correctness andrew gelman

Rationally Speaking #147 - Andrew Gelman on "Why do Americans vote the way they do?"

Sped up Rationally Speaking

Play Episode Listen Later Jan 3, 2021 50:49

There are two contradictory stories about politics and class: On the one hand, that the Republicans are the party of the fat cat businessmen and the Democrats are the party of the people. And on the other hand, that the Republicans are the party of the salt-of-the-earth Joe Sixpacks, while the Democrats are latte-sipping elites. In this episode, professor of statistics and political science Andrew Gelman shines some clarifying light on the intersection between politics and class in America, explaining what the numbers really show. He and Julia also cover the question, "Is it rational to vote?" Sped up the speakers by ['1.17', '1.0']

america americans vote republicans democrats sped andrew gelman rationally speaking

How Stitch Fix Relies on Data Science to Build the Perfectly Personalized eCommerce Experience

Up Next In Commerce

Play Episode Listen Later Dec 31, 2020 54:01

Style is a very personal part of what makes someone who they are. The way you dress is a reflection of who you are or who you want to be, and what speaks to you may be totally foreign to the next person. Knowing all of that, it’s understandable if you believe that something as personal and experience-driven as style could never be boiled down to data points or plugged into an algorithm. But… you’d be wrong.At Stitch Fix, a combination of human stylists, powerful A.I., and behind-the-scenes technology has created a winning model that delivers a personalized online shopping and styling experience straight to clients’ homes. A powerful data science team is one of the key reasons that Stitch Fix has been able to launch its valuation into the billions. Stephanie Yee is the VP of Data Science at Stitch Fix, and on this episode of Up Next in Commerce, she explains all the ways that data and technology are being put to use to create the best customer experience possible.Stephanie describes how technology like GPT-3 is going to finally make seemingly unimportant data consumable to a consumer audience, and she explains how an event like COVID-19 can impact your backend models and what to do to adapt in that situation. Plus, she gives tips on how any ecommerce operation can go about building a data science team, and the soft skills to focus on when hiring talent. Main Takeaways:Asking the Right Questions: The most important skill a data scientist can have has nothing to do with technical prowess. It’s about having the ability to frame a problem and then ask and answer the right questions. Encourage your team or new candidates to pump the brakes and reevaluate the “why” behind the question they are trying to answer or the problem they are trying to solve. Making The Indecipherable Easily Digestible: With the shifting demographics, and older generations now becoming more comfortable shopping online, tools need to be created to ingest and answer long-form questions in a way the consumer connects with. Technology like GPT-3, which is the most advanced language model to date, has the ability to do just this. Tune in to hear how! The Quick Change: Deploying algorithms and A.I. in conjunction with human resources/industry experts is critical for organizations to be able to adapt to big changes in a market. COVID-19 had a drastic impact on models that were trained on pre-COVID data. Should you scrap your current model and start over? Or build on what you have? Stephanie says a little bit of both.For an in-depth look at this episode, check out the full transcript below. Quotes have been edited for clarity and length.---Up Next in Commerce is brought to you by Salesforce Commerce Cloud. Respond quickly to changing customer needs with flexible Ecommerce connected to marketing, sales, and service. Deliver intelligent commerce experiences your customers can trust, across every channel. Together, we’re ready for what’s next in commerce. Learn more at salesforce.com/commerce---Transcript:Stephanie Postles:Welcome back to Up Next in Commerce. This is your host, Stephanie Postles, co-founder of mission.org. Today on the show, we have Stephanie Yee, the VP of data science at Stitch Fix. Stephanie, welcome.Stephanie Yee:Thank you. I'm excited to be here.Stephanie Postles:Me, too. I know it's going to be a good interview when there's two Stephanies, but I'm slightly worried about how the transcript will look. Like who's saying what? Who sounds smart? I'll just take all your quotes and pretend they're mine.Stephanie Yee:Perfect.Stephanie Postles:So tell me a little bit how long have you been at Stitch Fix for?Stephanie Yee:I've been at Stitch Fix for almost four years. Yeah, four years in January.Stephanie Postles:Well, tell me a little bit what does the role of the VP of data science look like day-to-day?Stephanie Yee:Yeah. If I have to think about it, being the VP of data science, it really comes down to maximizing the value that the data science itself and the team can bring to the company, like how do we really get the full promise of an algorithm's approach to things? I think as you guys probably know, Stitch Fix is really thinking about how do we help people find what they love and how do we use data science and human expertise to do that? So the types of things that I think about in service of that are things like what are new opportunities that we haven't really discovered yet? And that's been pretty exciting over the last four years.Stephanie Yee:I think another area that I think about a lot is like what's the right almost interface between data science and data scientists and the business partners. So this is if we have data scientists working with the design team, or the product team, or the marketing team, or even executives, what's the place where data scientists can contribute the most? And also, just being really intellectually honest, like what's the place where it makes sense for others to take over? And then obviously, the last part of my job is to really create an environment where the team can be motivated and fulfilled in doing things that bring out the best to each of them.Stephanie Postles:That's great. So it would be great to dive a bit more into Stitch Fix. I know what it is because I'm a customer, but I think a lot of people may not know exactly what it is or all the things that go on behind the scenes to get the pretty box on your door. So could you explain what it looks like, what is Stitch Fix from a high level, for anyone who doesn't know, and then what goes on behind the scenes to create the company that it is?Stephanie Yee:Yeah, so Stitch Fix is a personal styling company. And at the core, we use both data science and real stylists and their expertise to help people find what they love. If you think about unpacking that, it's really about understanding... or from a data science perspective, it's really about understanding a client's needs, as well as being able to set the stylist up for success. The core of Stitch Fix, the way that it shows up is in a box of one or more items and clients are able to try it on, they're able to send back what they don't like and really just keep what they really love.Stephanie Postles:Tell me how do you go about making sure that you give the customer the exact outfits they would like or refine that process to where maybe the second or third time you've nailed it? Because for me, at least when I am getting the outfits, I'm like, "The first time, maybe like one thing was off or something," but then after that, it's like, "Okay, now, this stylist knows me, or this algorithm knows me." So how do you refine that behind the scenes?Stephanie Yee:Yeah. I think that that's a great question. I think a lot of it... I mean, as a data scientist, like I always think about the data that we collect and what's available, and this comes both from what clients tell us as well as what we're able to infer, so a really interesting example of this, and this is where you had mentioned like, "Okay, there might be one item off at first and the algorithm really learns over time," we really think about things in terms of the ability to say like, "Okay, what data do we have now?" And with the stylist, the stylist is incredibly important throughout the client's life cycle. With the stylist, like what's the right thing to be sending right now? And in response to feedback like, "Oh, that item that didn't really work out for whatever reason," we're able to respond to that.Stephanie Yee:I think a really interesting example of the approach that Stitch Fix takes, or rather one of the interesting things about Stitch Fix is that we're thinking about this and we're thinking about a purchase experience in terms of soft goods. So if you think about the way that ecommerce really started off, or at least as I recollect it, it was like comparison shopping sites where you were looking at like how many megapixels do you want in your digital camera. And a camera, those are very easy to compare because it's like, "Oh, it is three or it is four." Whereas with what I think of as soft goods, there's so many different variants on like a V-neck top that it's almost a little bit overwhelming.Stephanie Yee:And then on top of that, a lot of the typical searching and filtering is not really going to get people there, just because what might be a great top, even if it's the same aesthetic, what may might be a great top for you might be not as great for me or vice versa, just because it's like, "Oh, you know what? I really need things that are machine washable, or I have very narrow shoulders or something like that." So Stitch Fix is really trying to distill a lot of these things that are ultimately very difficult to categorize into what we would call a latent space, but really to say like, "Okay, we have something like style." Style is not what lunch table did you sit at in high school, it's really a form of self-expression. And because people are so different, we need to be able to use data science to quantify where people are on a spectrum versus what category they're in. To handle this like-Stephanie Postles:How do you even encourage people to get maybe the feedback that matters? Because I'm even thinking like if I were to get a shirt and I'd be like, "Well, it doesn't fit," I know you're probably behind the scenes like, "Well, why? What part doesn't fit? What don't you like?" How do you encourage someone to tell you what you need to know to then send them something better?Stephanie Yee:Yeah. That's a great question. So what we found, I think that the motivation to give us feedback is actually just an inherent part of the service. I think a lot of people they'll like... When I've styled people and maybe I've missed the mark, people will say, "Oh, you know what? You didn't get it right the first time, but here's more what I was looking for." If you think about it as a relationship, it's not a transaction where you walk into the store and you say like, "I'm happy or I'm sad." It's relationships and relationships are predicated on that back and forth. So it's really a phenomenal percentage of clients that leave feedback on a fix. It's something like 85%.Stephanie Postles:Wow, that's great.Stephanie Yee:It's just like an intrinsic part of the relationship just because we do frame it as a relationship.Stephanie Postles:Yeah. I think having that stylist there really is what forms a human connection, where you're like, "Well, this is..." Of course, there's a bunch of machine learning and algorithms behind the scene, but there's a face here, a human who's actually approving this style and making sure it's perfect for me. And you instantly feel that connection and you don't want to let your stylists down [crosstalk] get that feedback.Stephanie Yee:Exactly, exactly. And similarly, the stylist doesn't want to let the client down. So there's that level of trust that gets established. And from there, I think, a lot of the desire to say like, "Hey, this had a fit issue for me because it was too long or something like that." There's just something that's special there inherently. And then on top of that, we obviously do encourage clients to give us feedback, like we'll give them a nudge. But we're certainly not the type of company that has to like... They'll come to us rather than us having to really force the issue, will say.Stephanie Postles:Yep. What are some of these subtle nudges that you give that aren't annoying, but then encourage the person to give you the information you need to help them. I think a lot of brands struggle with that, where they either don't follow up at all sometimes if they want feedback, or they do it too much and you're like, "Whoa, chill." How do you guys get that right blend?Stephanie Yee:I think that there's two parts to that. One is saying like, "What's the right number of times to be asking or to be reminding really because it's less on asking?" It's just more like, "Hey, if you want, you can leave feedback and there's someone on the other end who's going to be really thinking about it and responding to it." I think it's figuring out like what's the right time to tell people, and it's really like when would this be relevant to someone? I think that there's some other aspects where it's like what's the right time of day to reach out to someone? And all of these can be distilled down into data science problem or data science opportunities. I really find that to be really interesting. I think that there's another aspect, which is that the clients do come back to the app and come back to the site, even without looking to transact. Once they're there, then it's possible to be like, "Oh, by the way, did you want to..." Just making it really easy and lowering friction to giving feedback. That's another way that we're able to implicitly encourage it.Stephanie Postles:Yep. So with all this feedback coming in, it's a lot of natural language that you're probably getting, or is there any tech that you're excited about right now to help you categorize it? Are you looking into GPT-3 or anything new this year that could help solve that problem when people are just giving you probably long paragraphs of like, "Here's the things that aren't working for me," and they're just putting in terms that you're like, "Okay, I can actually build any database at this."Stephanie Yee:Yeah. I think it's interesting because I think some of the unstructured texts or data generally that might be, I would say, overwhelming to someone like you or I. Computers are actually quite good at processing it. So I think GPT-3 is really incredible, sort of advanced in the way that we're thinking about the opportunities that come from natural language processing. So I think the team is really actively thinking about like what's the right way to bring that into the client experience? We certainly want our stylists to continue to be proactive and like a central part of that relationship. And we're actually trying to figure out like, "Okay, how can we actually bring the stylist forward even more?" But I think the way that I would look at it is I actually love it when there's a large corpus of data, will say, just because there's quite a bit of things that one can infer or pull out of that that would be otherwise a rather arduous task for a person like you or I.Stephanie Postles:That's great. Earlier, you were saying that the team is looking at how to maybe utilize GPT-3. And it can be for the overall industry, not just Stitch Fix too. Is there anything where you're like, "I could see this really impacting ecommerce in this way," because that's the one area that I've been trying to look into it? I can see all the things that you can do with it, from not having to code things and writing books and stuff. How could it actually impact ecommerce or data science or behind the scenes?Stephanie Yee:Yeah, I think that's a great question. I would say that if you think about... GPT-3 is a really great way to translate information into the format that people are used to absorbing information in, which is text. I think that it's especially important going back to the like you can't take a shirt. The specs of a shirt are not particularly helpful to a shopper. They can be helpful to a computer, but it's like, "Okay, the sleeve is 13 and a half inches, like who cares?" And GPT-3 is able to almost add in a way that would have been incredibly difficult before. It's able to translate some aspects of an item into what that actually means in someone's everyday life. So it's not like, "Hey, we could show you a table of information where it says, 'Here's the sleeve length.'" But it can be more like, "Oh, you know what? This shirt is going to hit your elbow and it's actually going to drape a little bit."Stephanie Yee:And because there's so much clothing out there and it's all slightly different in its own way, even if it's once again, the same aesthetic, same color, everything, we're able to bring that to the fore for a massive amount of inventory. So that, I think, gets me really excited. I think another thing that's really promising about something like GPT-3 is it'll let us... yeah, it'll really let us customize an experience to a client using a format that... and move beyond tables of data into information that might be more relevant or easier to absorb.Stephanie Postles:Oh, that's great. Yeah, that takes it to a whole new level. I think about right now when I'm shopping around and it shows, "Okay, there's this model and she's 5'9 and 135 or whatever it is," I'm like, "Okay, I could see maybe how something would fit if that's like a similar person to me." But that takes it to a whole new level and saying, "All right, Stephanie, this is going to be bad, you get your elbows and it's going to be very short on your waist," and just putting it in a contextual term where I'm like, "Oh, it's fixed. You all know me. Thanks for letting me know."Stephanie Yee:Yeah. And if you think about style and aesthetic, there's almost something... Style is a form of self-expression. And describing style in terms of only numbers is quite limiting. If you look at the way that people will describe clothing, and it's always really interesting to say like, "Okay, fashion week happened. What are they saying about what's being shown?" It almost becomes poetic in that level of abstraction. And I think that that's something that that language is much better at doing, or images even are much better at doing than just numbers and texts.Stephanie Postles:So the thing I was just thinking about... I mean, you guys have all these models running and algorithms behind the scenes and you have really large amount of data. How have your models changed? I'm thinking about like pre-COVID models, [crosstalk] probably around work and work clothes, and I want to look nice and heels. And then now, it seems like all those models probably had a big shake up because now it's, "I want athleisure and I want sweat pants and comfy hoodies." How have you guys models changed and what are you doing to adjust them, or what should brands be thinking about with adjusting their historical models that are probably wrong?Stephanie Yee:Yeah, I think that's a great question. It was funny actually. In April, one of the data scientists posted in Slack and he was like, "Oh my gosh, like all of the experiments that we're running, we're just going to have to start over." I think that there was a lot of stress behind that statement. And obviously, we're not starting over, but we're starting from a place where the data has changed. And the really wonderful thing about an algorithm and about being able to really take advantage of technology is that they can adapt much, much faster than a person. If we only had the styling team, it can take a little bit of time to figure out like how do we... If we're learning something about like COVID trends, how do you train a team of thousands of people to be on top of everything that is there, in addition to letting them style each client individually?Stephanie Yee:So what's really wonderful... and COVID was a fascinating situation because it's like, "Okay, all of the..." There was a tremendous amount of work that had to be done to say like, "Okay, given a pretty big step change in the way that both like the world writ large as well as the way that people are thinking about shopping and shopping online, how do we adapt things to that?" So there was quite a bit of work to do that across the board. And then on top of that, it was easier than it would have been if we hadn't taken a data science approach, just because so much of the models are designed to change. Some of our algorithms they'll be like, "Okay, this is just going to be updated every week just because it needs to be."Stephanie Yee:I would say that in terms of COVID specifically like... And a lot of it, we're sitting there and we're saying like, "What are people like?" We'll have conviction in where we think that the market is going. With COVID, it was like, "Okay, you know what? Everyone can anticipate. If people have to stay at home, then they'll have to work from home and maybe they won't feel a need for as formal closes as they normally would." But what was interesting, and this is just from how things unfolded from a data science perspective, we actually had... One of our data scientists was a former epidemiologist. So when we were trying to figure out like, "Oh gosh, the world has changed, like how much merchandise should we buy a year from now," she was actually able to contextualize a lot of the news.Stephanie Yee:As a company, we were able to come to what ended up being a pretty reasonable, I would say, assumption about the world and then to go forward and say like, "Okay, overall, how much should we buy?" And then within that, it's like, "Okay, how are consumer tastes going to change?" We can lock down that merchandise. I think the merchant team did a really great job responding to that. Within that, we can make sure that the clients who are looking for working from home clothing versus something else they can actually get it. I think in terms of general trends, I think it's like a 10X increase in requests for working from home clothing. Definitely, a shift out of formal work wear and into more like casual and everyday styles.Stephanie Yee:I think athleisure, those purchases have accelerated quite a bit. With Stitch Fix, because we sell actual items, the merchant team had to do a tremendous amount of work to really anticipate that. And then the styling team is able to make sure that those items get to the right people. Because if suddenly we started to only send out leggings, that's not really going to work for many of our clients who just need to make sure that the people who are looking for athleisure can get it.Stephanie Postles:Yeah. That's so smart having someone who understands that industry. I feel like there's more room for brands to partner with industry experts like that to help them build their models. Because oftentimes, it seems like everyone is so focused on just, "This is our company model. Only the executives of the company can figure out what the future looks like." But by tapping into someone who has very different experience, [crosstalk] maybe what's happening, it seems very smart.Stephanie Yee:Yeah. One of the things that I find to be really fascinating and amazing about Stitch Fix is the way that the executives... like for executive decisions are able to take advantage of the data science capabilities that we've built. And you almost get to this like the core question here, and this is almost... it gets existential, like is how do you handle uncertainty? For me, I'm like, "Okay, this is why I want an executive with like 20 or 30 years of prior experience because some of these questions are genuinely hard."Stephanie Yee:I want to arm them. Given the data available, the task of a statistician is to really squeeze out as much information as possible and to say like, "Okay, guys, here's what we can know, here's what we can't know. And the part that we can't know, to the extent that it's incredibly important to have a decision or a point of view on that, that is truly human judgment." So the executive version of that, I find to be really interesting and there's many versions of that throughout the company with the stylist, with the product team, with the marketing team, with the merchandising team, everyone.Stephanie Postles:That's great. So when thinking about updating the models and algorithms, would you suggest that a company rebuild from scratch, or should they update a current model to kind of pivot a bit? Because I guess when I think about updating a current model, I worry that there's so many things built into it after the fact and the algorithm just runs away on its own and people are like, "I don't really know what's driving it anymore," versus starting over again.Stephanie Yee:Yeah, I think that that's a great question. There's a couple of different aspects to it. Generally, we'll think, "Okay when you have a..." because a model is really expertise in how to use data. So if you find a model that seems to fit the world very well, then you will want to continue to improve it. If fundamentally the world responds quite well to a random forest, or we get very good predictions out of a random forest, then there's no need to change it just because, but there's opportunity to improve on that. Now, with that said, as research is continuing on different methods, people are going to try different methods. But I would say that you definitely want a mix of both because it's both the method and the tuning of that. It's both the type of model that people will think about as well as the tuning of that model or adding new variables to the model or something like that that we want to do.Stephanie Yee:So to give you a concrete example, like with COVID, we have a demand forecast. The demand forecast is really modeling client behavior and it's really being able to give the merchandise team and the executives and the operating partners visibility into like, "Okay, what's life going to be like a year from now and how should we plan?" When COVID happened, everyone's like, "Oh my gosh, the world is very different." But what was great was we were able to say, "Okay, here's some assumptions that we have. We can update those assumptions, but we've got several years of work into the capability itself. And the great news is that we don't need to start from scratch because things have been built in a way that can adapt."Stephanie Postles:Yeah, that's very smart. When thinking through your demand forecast, are you guys forecasting that the world will eventually return back to pre-COVID, or do you think it's a new normal and now people are going to continue working at home indefinitely and keeping it adjusted? How are you guys forecasting the future of apparel?Stephanie Yee:Yeah, I think that's a great question. I would say that there are certainly things that are very large shifts and there are other things that are just probably going to stay the same. I would say that it's a blend of the two. I certainly don't think and I certainly hope that we're not going to be working from home forever.Stephanie Postles:Yeah, I hope not.Stephanie Yee:Exactly, exactly. With the vaccine coming out and just how effective the vaccine seems to be, I think that we will be returning to... There's some things that are going to fall back into place. There's some things that frankly have already fallen back into place, and then there are other things that the company is really leaning in to take advantage of, so definitely a mix.Stephanie Postles:Yeah, yep, I agree. Have you seen any different types of consumer buying behaviors around what consumers are expecting now that more people are at home, they have more time to try things on? Have you had to adjust how you interact or work with your consumers during this time that was maybe different than COVID or pre-COVID?Stephanie Yee:Yeah. I think as I mentioned, there's definitely a difference in what it is that people are looking to buy. I think another thing that has been really exciting is that I know quite a few new shoppers, people who have never bought anything online before suddenly they're like, "Oh, shoot, all the stores are closed. I now have to try this new channel." So we're seeing people who aren't even used to a traditional way of shopping buying things. I think that's been really interesting because that behavior, as you can imagine, can be quite different. So it's great that the business is able to respond to that and-Stephanie Postles:Yep. It seems like there's a whole new demographic market that is opening up now that a lot of ecommerce companies are going to be able to have a lot of opportunities with. I'm thinking about Stitch Fix, my mother-in-law, who's almost 70, when they came back and told me she... I had never told her about you guys and I don't think she would actually ever do that. And she's like, "I ordered from this company, they picked things out for me, it fits perfectly." And I'm like, "Are you talking about Stitch Fix?" And I was genuinely surprised, but she found out about it on her own, went forward, bought it, worked with a stylist, and got her box. It just made me think about how many opportunities are opening up with this new group of people who never were probably comfortable with buying online before. But now, they're forced to it and it's now becoming normal for them.Stephanie Yee:Yeah. I love that story. That's wonderful. I think what's interesting too is that folks like your mom or my mom where they're not actually as used to buying online, they're more used to going into a store, so they're actually more used to being able to talk to someone. Whereas like my friends are like, "I don't want to talk to a human being."Stephanie Postles:Yeah, don't call me. Don't look at me.Stephanie Yee:Just text me. Right. Don't leave a voicemail, that doesn't work. Right. But the folks who are trying something online, they're used to a store. And Stitch Fix like the gap between some of these department stores where you do have a person and the department stores online presence is quite uncomfortable. So if you have Stitch Fix, obviously you're not in the store, but you get to try things. You get to work with a person, you get someone who's actually there to help you. I think in some sense, it's actually a more natural entry point, especially if folks aren't used to the current paradigm of shopping.Stephanie Postles:Yeah. How would you advise a company to be able to not only continue to focus on their traditional consumers that they're used to, but also lean into that new group of people because it seems like you would have to have very different messaging? Like you were just mentioning, some people like myself and you are like, "Just text me, do not leave me a voicemail. Don't try and call me, I'll decline it," whereas this group, you have to have a whole different mentality. Your customer service team probably needs to start calling people and doing things very different. How would you advise a company thinking about this, who wants to maybe connect with both of them, their current customers and the new ones who are now coming on the market?Stephanie Yee:Yeah, I think that's a great question. So the way that I would think about this is, first off, you have to come up... Let's take the messaging example. You want to think about what are the different messages that are going to be resonating with consumers? And then the second is how do you get the right message to that consumer? In terms of what will resonate, I firmly believe... There's a very interesting opportunity for interaction between design and data science and user research and things like that. Data science can contribute, but ultimately the messaging strategy is one that is the overall messaging strategy. You can try many different variants, but the overall strategy is one that is a judgment call.Stephanie Yee:And then machine learning is wonderful for being able to say like, "Given this message, or given this client and given a universe of messages, how do I make it so that the client can see the most compelling one and really understand on their terms what it is that we can offer?" I would say this is an area where you definitely need both art and science because messaging is so incredibly important and strategic. So it's working with the marketing team, it's working with the design team, and then the data scientists can really help figure out where should that message be delivered? How should it be delivered? What is the right way to make it land with the client?Stephanie Postles:Oh, that's great. So we're putting together this end of year commerce article about 2021 trends. And this is one thing that we're talking about is how much the over 55 demographics spend. And they spend twice as much as millennials. And I think I saw, let's see, 10,000 baby boomers are going to turn 65 every day until 2030.Stephanie Yee:Oh, wow, okay.Stephanie Postles:And then by 2050, the over 60s will account for 20% of all people globally. So when I started seeing these stats, I'm like, "Whoa, more people need to focus on this demographic." Oh, and then another one, the 50 and older crowd has a lot of spending power. And if you put it in terms of GDP, it would be the third largest in the world.Stephanie Yee:Oh, wow, okay.Stephanie Postles:US is 21 trillion, China's 14 trillion, and then Japan is 5 trillion. And this is where the people they spend 7.6 trillion in 2018. To me, I'm just seeing all these opportunities that are being missed right now. I'm like, "What the people be doing?"Stephanie Yee:Yeah, no, I think this is a wonderful group of folks. Within the tech industry, I would just say especially there does tend to be a focus more on millennials and things like that. I think the great thing about Stitch Fix is that we are... And oftentimes I think some brands they'll sit there and they'll say like, "Oh, our target demographic she is between 25 and 39. And after that, she's not us." I think with Stitch Fix, we're able to say, "You know what? We're not going to categorize you into one group or another, we're going to serve you where you are. And with personalization, we are able to..." I completely agree with the stats or the information that I have on how that generation of clients interact with Stitch Fix is very, very consistent with some of the numbers that you had described. So it's a really wonderful group of people who are thinking about their personal style, and I do agree it's folks who, I think, tend to be served a little bit differently, really at the retail industry's loss.Stephanie Postles:Yep. Yeah, I agree. How would you go about getting the right data to then be able to craft the personal message then? For Stitch Fix, it does feel a little bit easier because you can ask things like age and a bunch of other questions and they're like, "Well, they're styling me." But for a lot of other brands, if you were to ask age, they'd be like, "What?" How would you advise other companies to be able to get enough information to then be able to personalize a message like that?Stephanie Yee:Yeah. I think that there's a couple of different ways to do it. And a lot of it really is around the marketing and design toolkit. Because ultimately, when you're coming up with messaging, you don't want to say like, "Okay, this is the messaging for folks who are 50." I'm an old soul, so maybe I'll just really respond to that myself. A lot of this is just a strategic question. So data science can play some role where it's like, "Okay, based on what we know, people tend to respond to X, Y, and Z." But really, if you want to be looking forward, it's less like what've people responded to in the past?" You definitely want to take that into account, but it's more like where are things going in the future, especially at a time when things are changing so rapidly?Stephanie Postles:Yep, yeah. That's why I'm also excited about being able to ingest the sentences that people are asking the customer service reps or putting in the search bar because I think that alone could tell you who someone is just based on how they say [crosstalk 00:32:09].Stephanie Yee:Oh, absolutely. Absolutely. Yeah, I agree. I think that the notion of being able to have more conversations with people is something that I think is incredibly exciting and it does allow for a level of, I would say, flexibility of expression, especially once computers can really respond to that.Stephanie Postles:Yep. So when thinking about building up a data science team, what are your first steps? How would you tell a brand to think about it to be able to build it up in an efficient manner, where it's answering the right questions, you have the right goals in mind? Because when I think about data science from different companies I've worked at, some people are called data scientists when they're really a BI team [crosstalk] called data scientists. And then you have marketers who are also data scientists. So like how do you [crosstalk 00:33:26]?Stephanie Yee:It's certainly become a loaded term. It's funny because in recruiting, it can be incredibly frustrating like, "Well, this LinkedIn search is not very helpful." Yeah, that's the first thing that I would say is if people are thinking about like I need to build out a data science team, searching on the term data scientist is probably not going to be the most efficient way to get there. I think probably the step one that I would advise people to do is to really think about what role do you want data science to play and where are the areas that you see as high value? And this can be a little bit of a hard question because without... in the same way that I'm not a 100% familiar with a merchant's toolkit or a designer's toolkit.Stephanie Yee:If I, as a data scientist, look at a problem, I can be like, "Oh, this is something that can be very easily solved with the machine learning." It's hard for folks who don't have that background to know that, but really thinking about like what is the strategic problem that people are trying to solve? With data science, I'm very supportive of making it like a core... like figuring out how to have in-house data scientists focused on the core problems of the company. So it's like what are the core problems of the company? What role would you want data scientists to play within that? I think one of the things that's wonderful about Stitch Fix is that data scientists are really expected to take a leadership role. And this can be incredibly exciting for some folks and it can be just not really interesting to others.Stephanie Yee:So figuring out like, "Okay, if you want data scientists to play a strategic role like, A, what's the core of your company, B, can you hire people who are inclined to really step up and to contribute to that strategy, and then C, how do you set them up for success," I think... And when I've talked to companies, some people will say, "You know what? We're really about logistics." And it's like, "Oh, actually, there's a subset of data science," where they're really thinking about operations research, they're really thinking about warehouse efficiency, supply chain and things like that.Stephanie Yee:And if people are really thinking about demand forecasting and logistics and fulfillment, that's a great tack to go for Stitch Fix. A lot of it is around the warehouse and the fulfillment side of things, those folks who are doing wonderful work and it's all in service of a very specific type of client experience or being able to provision a specific type of client experience. So we have folks who are working on the warehouse side of things, but then we also have folks who are really thinking about like, "How do you work with a stylist to help people find what will really help them or what will really bring them joy?"Stephanie Postles:I got it. If a company doesn't really... They know some of their problems and they know their operations, but if they don't know data science, how would they know what they can solve, or how would you recommend, like should they go and talk with the company or a mentor or advisor who understands that area to do just what you just did with me of like, "Oh, of course, you can put it in logistics and you can put it on your website or here"? How would you tell someone to move forward if they don't know what they don't know?Stephanie Yee:Yeah. If you tell me what is most important to your business, then I can help you figure out like what are the data science opportunities there. And sometimes data scientists may not be the most important input to that, at which point then there might be alternative areas to invest.Stephanie Postles:Yep. What kind of skills would you be looking for when you're hiring a data scientist team, or what are some of the emerging skills too that you're like, "We weren't looking for this four years ago, but now it's something that's very much in demand"?Stephanie Yee:Yeah. That's a great question. I would say the skill that seems to be more and more in demand, and this is something that I think from the early days Stitch Fix had good intuition that this was important, is around problem framing. Like a data scientist, we need them to have a good understanding of statistics, oftentimes machine learning, computer programming, sometimes software engineering. But really, the core thing that we think about is like, "Can they frame a problem and can they... How do they think about problem framing?" Because what will often happen, and this is a pattern that I've seen in other places, is people will very valiantly answer the wrong question. And it's not their fault that they're answering the wrong question, it's just the wrong question was asked.Stephanie Yee:So what we really encourage folks to do and what I think the most effective data scientists do when they're empowered to do so is if people pose a problem to solve, it's actually okay to say like, "Okay, let's take a step back. Let's dig into this a little bit and figure out like is this posed in a way that can lend itself to the full suite of potential solutions?"Stephanie Postles:Got it. So if you're interviewing someone, how can you test that when you don't have much time with them? What kind of questions can you ask to see are you able to actually ask the right questions to figure out what the problem is without going down the wrong path right off from the start?Stephanie Yee:Yeah. That's a great question. Oftentimes there's two ways to do that. One is to say like, "Okay, tell me about a time when someone has posed a very vague business problem and how did you think about refining what it was?" I think that that's one angle. And then another angle that I will bring to the table is I'm thinking about this type of problem, how would you help me? How would you think about it? And just really making it into a discussion because what you're really looking to assess is how do people think? And I will say interviews are not... When you only have 45 minutes with someone, or you have six people with 45 minutes each with someone, you don't get nearly as much data as you'd ever want to. So when I think about it, I just want to have a conversation and see how people think and what connections do they make. If something is framed in a way that merits revision, how do they go about figuring out what that revision might be?Stephanie Postles:Great. So in an industry that's changing so quickly, how are you staying on top of new trends and tech? Are you subscribing to a bunch of newsletters? Are you listening to podcasts? What do you do to stay on top of the data science field?Stephanie Yee:I think that that's a great question. I do subscribe to newsletters. There's a couple of blogs that I really like as well. I think Andrew Gelman is a professor, I believe, at Columbia. He has some wonderful work. Susan Athey is actually another researcher at Stanford who I think is absolutely wonderful. She thinks about causality. So this is like what actually causes another thing. And she thinks about machine learning techniques that can... One of the areas of her research is thinking about how machine learning can contribute to that field. I personally like to stay closer to some professors that I particularly admire. And then also the great thing about Stitch Fix is that everyone has a different set of passions and interests, will say, as well as a different background. So when people are coming across a lot of different methods or papers, there's a wealth of different conversations going on. So that's another great way to stay on top of things.Stephanie Postles:Yeah. I found it really helpful when I dive into certain trends. Like every week I'll pick a new piece of tech or a new trend or something just to see what it's about. And then I start to realize how many new things I'm being introduced to and new people on Twitter that I'm following and new ways to solve problems, like at a media company with podcasts, where I'm like, "Whoa, I never thought about using that. But now that I've read about it, I can think of 1,000 ways to maybe implement it, or I have a whole new model in place or a business model idea based on just very things that are not a part of maybe the media industry or something."Stephanie Yee:Exactly, exactly. I love finding metaphors in one area that tend to work in another. Being able to abstract between things is such a source of insight. I agree with that.Stephanie Postles:Yep. So where do you see the future of data science in ecommerce headed? How do you see that experience playing out in the next five years or so? What does it look like, or what does it feel like?Stephanie Yee:Yeah, I think that's a great question. I think that the future of ecommerce is really one where you have a more personalized experience. I think that as we've discussed, data science is an incredibly important input to that in being able to really fulfill on that promise. I think that data science can also help retailers make better decisions. I see a lot of promising growth on that front. I think for retailers who are particularly fulfillment or operationally focused, there's some really wonderful sort of... I think Amazon is really leading the way in the direction that that side of things can go.Stephanie Postles:Mm-hmm (affirmative). I see a lot of companies probably looking to this field, especially after all their models and plans started breaking last year, trying to figure out what can I get ahead of this next time? There's going to be a next time of something and how can I get ahead of that and start seeing the early indicators and maybe be able to be more agile with adjusting forecast and supply chain and all of them.Stephanie Yee:Yeah, absolutely. I think that that level of agility is something that I'm very proud of that Stitch Fix has. And part of it is because we're able to use data science, not only it's like, "Okay, we can update this model relatively quickly compared to others, or we can figure out how to take into account in the past but not too much," but then also in the ability to help executives think through different scenarios. Because ultimately, we can use data to do some things like we need to executive input on other things.Stephanie Postles:I always love a good data story. Are there any stories that come to mind that either the data shows something that was wrong or it was funny, any of your favorite data stories that you think about from time to time?Stephanie Yee:Yeah, I think that's a great question. The one that I find to be quite endearing is... We have this notion of latent style. And this is rather than saying like, "Oh, here's the lunch table I sat at in high school," it's where within the broader realm of style do you sit? And one of the early hypothesis, it was like, "Okay, we have this sense of the types of clothes that people will like, and we can show pictures of them. But we should figure out how to articulate this to a stylist." So there was some work done to say like, "Hey, here's a set of clothes and here's another set of clothes, named them." So you could be like, "Oh, this is casual and preppy, or this is boho and edgy or something like that."Stephanie Yee:We basically ask people to annotate collections of clothing based on how they would describe that aesthetic. What was wonderful in a sense was all of the... There wasn't really that much consistency between what people were saying. I think sometimes people are like, "Oh, this is a problem." I was like, "No, this is great, guys. This is actually great because it proves that there's things that are there that are beyond categorization. I view self-expression and style as of them." Ultimately, when it was like, "Okay, now we need to express a client style to a stylist," a lot of it was just like, "Let's just show examples and pictures because we don't have the words for it."I thought that that was wonderful. In my mind, it really spoke to the value that Stitch Fix's approach brings to ecommerce.Stephanie Postles:That's cool. So that definitely shows that consumers on your side definitely can't be used from a... You do see annotation label or dataset type of aspect because they're all going to come back with, "This is preppy. Oh, no, this is boho. Oh, no, this is athleisure," and it wouldn't really work for you guys.Stephanie Yee:Yeah, yeah. It does become interesting. Because if you have something that is totally fashion-forward and wild, then nobody would certainly say like, "Oh, it's classic." So there might be a cloud around things. But it definitely does speak to like where is it that people can be most effective versus pictures versus something else?Stephanie Postles:Mm-hmm (affirmative). All right, cool. Well, let's move over to the lightning round. Lightning round is brought to you by Salesforce Commerce Cloud. This is where I'm going to ask you a question, and you have a minute or less to answer. Are you ready, Stephanie?Stephanie Yee:Sure.Stephanie Postles:All right. First up, what's the nicest thing someone's ever done for you?Stephanie Yee:For me personally?Stephanie Postles:Yep, personally.Stephanie Yee:Oh my goodness, okay, I'm going to need the minute.Stephanie Postles:Yeah, go for it.Stephanie Yee:I guess this is tangential to data science. But there's one point early on in my career. I started my career in management consulting and management consultants are an incredibly impressive bunch, incredibly good at dealing with uncertainty. They also have a very clear view for like what makes you successful management consultant. And there was a talk that one of the partners gave that stayed with me. I think of this as nice because it was quite informative in how I viewed the world and what I can do within it. He said like, "You know what? Every place that you work is going to try to put you into a box. They're going to try to categorize you and nobody really fits into a box."Stephanie Yee:So it's actually okay to say, "You know what? Try to find a job where you fit closely enough into... where all of the strengths that you bring to the table match the strengths that they're looking for." But it's actually okay to say like you have quirks and you're not always going to fit into the box. I think that that was really wonderful because it was like, "Oh, I need to find a job where the way that I think about things maps really well to the thing that the company needs." And then I also need to say like, "What are the things that I bring to the table that people might be like, 'Hmm, okay, that's interesting'"? And decide like, "Is that something that I want to develop about myself, or is that something that I want to say, 'You know what? That's just a strength that I have, or that's just an aspect of myself'"? I would say that that piece of advice was incredibly generous to give as well as something that was very valuable to me.Stephanie Postles:Oh, I love that. I'm really glad I asked that question now. It's a new one. So I always am waiting to hear if someone's like, "Ah, nothing," or something really great like what you just said.Stephanie Yee:Okay, I'm glad that that worked.Stephanie Postles:That was good. What is a trend or piece of tech that you don't understand today that you wish you did?Stephanie Yee:That's a great question. Let's see. A piece of tech that I don't understand today that I wish I did. I think on my list, I've been very interested to get in a little bit more into the weeds and how autonomous vehicles work. I've read at least like what The New York Times will say about Waymo or things like that. But I haven't gotten a chance to really read up on the literature. That's definitely been on my list in part because I think it's just a very interesting problem to solve. And I actually have some friends who are working on that problem, so I can probably just ask them, but also because it's something that is probably going to transform society in the next decade or so.Stephanie Postles:Yeah, I agree. What's your favorite data science book that you refer back to?Stephanie Yee:What's my favorite data science book? The core one that... And this one is not exactly readable, but it's quite nice to reference is elements of statistical learning. I would actually say that it's... Well, for some people, it's readable. For me, it's more of a reference book. But it's this wonderful collection of information put out by some professors at Stanford. I think that it's like a cornerstone of a lot of machine learning and data science classes.Stephanie Postles:What's up next on your Netflix queue?Stephanie Yee:Right now, I'm in the middle of The Crown.Stephanie Postles:Okay. I've had a lot of people say that.Stephanie Yee:Yeah. I hadn't gotten into The Crown actually until shelter at place... I sort of been like a elapsed inactive Netflix customer on and off throughout the years. But I had heard so much about it that I was finally like, "All right, I will sit down and watch it. It's really good.Stephanie Postles:Yeah. I started it and I'm excited to finish it. And I heard the next season's not coming out for like another couple of years or something.Stephanie Yee:I know. I was just like, "Oh, I should have waited to get into it until everything is done." But yeah, I think it's like two years.Stephanie Postles:All right. And then my last question, which is very important, how strictly do you enforce when people are writing up the term data? Do you use it properly, like the data shows, the data show? How strict are you with your team about use the word data properly?Stephanie Yee:I will say-Stephanie Postles:Very important [crosstalk 00:52:14].Stephanie Yee:I notice when people use it. I noticed people's grammar, including that. There are other concepts that I will become more passionate about than grammar necessarily. I think it's incredibly important, but I think the contents, like the true content and making sure that we're precise in certain other words is probably higher priority. I generally try to take a light touch with my team.Stephanie Postles:Okay, you're now stickler about it.Stephanie Yee:I do notice though. I have to filter it, will say.Stephanie Postles:I think you does it right. [inaudible 00:52:56].Stephanie Yee:I do, I do, actually.Stephanie Postles:I love that. All right, Stephanie, well, this has been a really fun interview. Where can people find out more about you and Stitch Fix?Stephanie Yee:That's a great question. The stitchfix.com website is probably the best place to find out more about Stitch Fix. I think in terms of myself, that's a great question. I do have a side project called RTD3.us.Stephanie Postles:I was looking at that. What is that actually? I saw it on your Twitter, but I didn't have enough time to jump into it.Stephanie Yee:It's just a side project. I was at a machine learning and fraud detection company at the time. Oh my goodness, this was probably like six years ago or so, maybe seven years ago when machine learning itself was just starting to like... It wasn't anything that people knew anything about. And a lot of the vendors out there would be like, "Hey, we have this super advanced algorithm, dah, dah, dah, dah, dah, dah." I found it to be a little bit annoying that people would market it as it's too complicated, you can't understand, ours is the best. And at the time, I was quite indignant because we actually had top-notch data scientists and engineers who did actually have something that was the best, but we were still trying to figure out how to market ourselves.Stephanie Yee:So I was like, "Okay, I want to sit down and I want to be able to explain machine learning and some of these more advanced statistical concepts to people who didn't take linear algebra in college." Very, very smart people who just decided to study something different because it doesn't have to be as difficult or as complicated as people make it out to be. There are some things that are incredibly complicated and wonderful and elegant, but you can distill something down to something that is accessible to a broader audience. So I worked with a designer/front end engineer, and we came up with something that really tries to explain some of these core concepts and to make it accessible to people who otherwise like... others are just trying to confuse them.Stephanie Postles:Yep, yeah. That's great. It reminds me... I mean, it's not very similar, but have you heard of Sideways Dictionary?Stephanie Yee:I haven't. I want to go check that out though. That sounds wonderful.Stephanie Postles:That's a dictionary and it uses analogies to explain technical terms. So very different than what you're talking about, but it's helpful because if you look at... Let's see, I'll look at API. The analogy is it's like the connectors on the back of your TV. They let you plug in a device from another manufacturer and both the TV and the device know what to do next. And the connectors are the interface that lets one machine talk to another.Stephanie Yee:Oh my goodness, I love that. This is actually something that I end up doing at work anyway. So I'll have to take a look at that. This is wonderful.Stephanie Postles:Yeah, check it out. I was looking at the about and I saw that it was created by Jigsaw. I don't know if you remember that. It's a group within Google. I think it's just one of their side projects that some of the engineers built. I'm like, "This is actually pretty helpful for me to understand technical engineering type terms."Stephanie Yee:Yeah, yeah, no, I think t's very easy to forget like what it was like to not know something. I think that for folks who can remember that, there's a great deal of empathy there and there's a great deal of desire to help people just understand technology in general. So I will definitely look at that. That's very exciting.Stephanie Postles:Cool. All right, Stephanie, well, thanks so much. Yeah, we'll have to have you back around since I feel like we have a lot of things we could keep talking about, but until next time.Stephanie Yee:All right. Thank you. This was great.

Millenniyule 2020 (12-14-20)

Luke Ford

Play Episode Listen Later Dec 15, 2020 166:21

00:00 Losing friends while streaming 03:00 Bill Barr resigns 19:00 Millenial Woes says I made a nasty video about him 32:30 What's Next for Trump Voters Who Believe the Election Was Stolen?, https://www.nytimes.com/2020/12/14/us/trump-voters-stolen-election.html 43:00 Andrew Gelman: 100 Stories of Causal Inference 51:00 When does peer review make no damn sense?, https://statmodeling.stat.columbia.edu/2016/02/01/peer-review-make-no-damn-sense/ 1:03:00 PORNHUB CANCELLED by VISA and MASTERCARD, https://www.youtube.com/watch?v=znUMjWdUzDY 1:05:00 Naked Yoga on Youtube, https://www.youtube.com/watch?v=qcNT_Ze4mck 1:10:00 Pornhub Just Purged All Unverified Content From the Platform, https://www.vice.com/en/article/jgqjjy/pornhub-suspended-all-unverified-videos-content 1:41:30 Multiple bowl hypothesis with Dooovid 1:45:00 Z-Man LIVE on corporatism and what it means for America, https://www.youtube.com/watch?v=_3CEoY_dgHs 1:48:20 Z-Man: Working families for Democrats are neither working nor families 1:54:00 Sportsmax 2:03:00 MILLENIYULE 2020: TRAD NEW'S PICK OF THE MILLENNI-YULES, https://trad-news.blogspot.com/2020/12/2020-pick-of-millenni-yules.html 2:03:00 Millennial Woes apologizes for WWII, https://youtu.be/YjvZDkia6kE?t=2011 2:05:30 A new approach to nationalism 2:07:10 Success is a good thing. 2:09:20 DANGERFIELD - ANOTHER "PATRIOTIC ALTERNATIVE" DEGENERATE, https://www.bitchute.com/video/gXMgbMT7PQtJ/ 2:14:50 JF REVEALS HOW HE GOT NICK FUENTES BANNED FROM YOUTUBE 2:26:40 JF Gariepy blames Jews for starting WWII, https://archive.org/details/181029garapeylyingaboutthejewsstartingwwiifromnowdeletedyoutubevideo MILLENNIAL WOES DESTROYS OWN BRAND BY GOING 1488, https://trad-news.blogspot.com/2019/01/millennial-woes-destroys-own-brand-by.html https://dlive.tv/LukeFordLiveStreams Listener Call In #: 1-310-997-4596 Superchat: https://entropystream.live/app/lukefordlive Bitchute: https://www.bitchute.com/channel/lukeford/ Soundcloud MP3s: https://soundcloud.com/luke-ford-666431593 Code of Conduct: https://lukeford.net/blog/?p=125692 https://www.patreon.com/lukeford http://lukeford.net Email me: lukeisback@gmail.com or DM me on Twitter.com/lukeford Support the show | https://www.streamlabs.com/lukeford, https://patreon.com/lukeford, https://PayPal.Me/lukeisback Facebook: http://facebook.com/lukecford Feel free to clip my videos. It's nice when you link back to the original.

america success stories dm losing world war ii jews democrats platform paypal visa conduct mastercard william barr degenerate naked yoga andrew gelman millennial woes

Presidential Polling and Constitutional Law - 12.13.2020

What Happens Next in 6 Minutes

Play Episode Listen Later Dec 13, 2020 121:08

Co-hosts Larry Bernstein and Rick Banks. Guests include Eric Kaufmann, Doug Massey, Andrew Gelman, W. Joseph Campbell, Larry Kramer, Mark Tushnet and Ganesh Sitaraman.

presidential polling joseph campbell constitutional law larry kramer eric kaufmann andrew gelman ganesh sitaraman mark tushnet rick banks

#29 Model Assessment, Non-Parametric Models, And Much More, with Aki Vehtari

Two for Tea with Iona Italia and Helen Pluckrose

Play Episode Listen Later Dec 2, 2020 65:04

I’ll be honest here: I had a hard time summarizing this episode for you, and, let’s face it, it’s all my guest’s fault! Why? Because Aki Vehtari works on so many interesting projects that it’s hard to sum them all up, even more so because he was very generous with his time for this episode! But let’s try anyway, shall we? So, Aki is an Associate professor in computational probabilistic modeling at Aalto University, Finland. You already heard his delightful Finnish accent on episode 20, with Andrew Gelman and Jennifer Hill, talking about their latest book, « Regression and other stories ». He is also a co-author of the popular and awarded book « Bayesian Data Analysis », Third Edition, and a core-developer of the seminal probabilistic programming framework Stan. An enthusiast of open-source software, Aki is a core-contributor to the ArviZ package and has been involved in many free software projects such as GPstuff for Gaussian processes and ELFI for likelihood inference. His numerous research interests are Bayesian probability theory and methodology, especially model assessment and selection, non-parametric models (such as Gaussian processes), feature selection, dynamic models, and hierarchical models. We talked about all that — and more — on this episode, in the context of his teaching at Aalto and the software-assisted Bayesian workflow he’s currently working on with a group of researchers. Our theme music is « Good Bayesian », by Baba Brinkman (feat MC Lars and Mega Ran). Check out his awesome work at https://bababrinkman.com/ (https://bababrinkman.com/) ! Thank you to my Patrons for making this episode possible! Yusuke Saito, Avi Bryant, Ero Carrera, Brian Huey, Giuliano Cruz, Tim Gasser, James Wade, Tradd Salvo, Adam Bartonicek, William Benton, Alan O'Donnell, Mark Ormsby, Demetri Pananos, James Ahloy, Jon Berezowski, Robin Taylor, Thomas Wiecki, Chad Scherrer, Vincent Arel-Bundock, Nathaniel Neitzke, Zwelithini Tunyiswa, Elea McDonnell Feit, Bertrand Wilden, James Thompson, Stephen Oates, Gian Luca Di Tanna, Jack Wells, Matthew Maldonado, Ian Costley, Ally Salim, Larry Gill, Joshua Duncan, Ian Moran, Paul Oreto, Colin Caprani, George Ho and Colin Carroll. Visit https://www.patreon.com/learnbayesstats (https://www.patreon.com/learnbayesstats) to unlock exclusive Bayesian swag ;) Links from the show: New podcast website: https://www.learnbayesstats.com/ (https://www.learnbayesstats.com/) Rate LBS on Podchaser: https://www.podchaser.com/podcasts/learning-bayesian-statistics-932588 (https://www.podchaser.com/podcasts/learning-bayesian-statistics-932588) Aki's website: https://users.aalto.fi/~ave/ (https://users.aalto.fi/~ave/) Aki's educational material: https://avehtari.github.io/ (https://avehtari.github.io/) Aki on GitHub: https://github.com/avehtari (https://github.com/avehtari) Aki on Twitter: https://twitter.com/avehtari (https://twitter.com/avehtari) Bayesian Data Analysis, 3rd edition: https://www.routledge.com/Bayesian-Data-Analysis/Gelman-Carlin-Stern-Dunson-Vehtari-Rubin/p/book/9781439840955 (https://www.routledge.com/Bayesian-Data-Analysis/Gelman-Carlin-Stern-Dunson-Vehtari-Rubin/p/book/9781439840955) Bayesian Data Analysis course material: https://github.com/avehtari/BDA_course_Aalto (https://github.com/avehtari/BDA_course_Aalto) Regression and Other Stories: https://avehtari.github.io/ROS-Examples/ (https://avehtari.github.io/ROS-Examples/) Aki’s favorite scientific books (so far): https://statmodeling.stat.columbia.edu/2018/05/14/aki_books/ (https://statmodeling.stat.columbia.edu/2018/05/14/aki_books/) Aki's talk on Agile Probabilistic Programming: https://www.youtube.com/watch?v=cHlPgHn6btg (https://www.youtube.com/watch?v=cHlPgHn6btg) Aki's posts on Andrew Gelman's blog: https://statmodeling.stat.columbia.edu/author/aki/ (https://statmodeling.stat.columbia.edu/author/aki/) Stan software: https://mc-stan.org/ (https://mc-stan.org/) GPstuff - Gaussian... Support this podcast

The problem with polling

Business Daily

Play Episode Listen Later Nov 13, 2020 18:29

Is the polling industry the real loser in the American presidential elections? Pollsters have come in for criticism that they misjudged President-elect Biden’s support, and did even worse in the state senate elections. Andrew Gelman, professor of statistics and political science at Columbia University explains why some of the errors were made. Zeynep Tufekci, associate professor at the University of North Carolina, Chapel Hill, argues that polling can have a distorting effect on democracy itself, changing how people vote or whether they do at all. Meanwhile, Anthony Wells of UK research firm YouGov explains how the polling industry functions outside of the electoral spotlight, and why political forecasts are just a small part of it. (Image credit: Getty Creative.)

american university president uk joe biden north carolina columbia university chapel hill polling yougov pollsters zeynep tufekci andrew gelman anthony wells

66 - Stuart Ritchie - Hype, Bias & Fraud in Science

Play Episode Listen Later Nov 1, 2020 95:30

Stuart is the author of Intelligence: All that Matters (2015): https://www.hachette.co.uk/titles/stuart-ritchie/intelligence-all-that-matters/9781444791808/ and Science Fictions: Exposing Fraud, Bias, Negligence and Hype in Science (2020): https://www.penguin.co.uk/books/111/1117290/science-fictions/9781847925657.html You can find some of his academic publications listed here: https://www.kcl.ac.uk/people/stuart-ritchie Follow Stuart on Twitter: @StuartJRitchie Further References Charles Murray, Human Diversity: The Biology of Gender, Race, and Class (2020) My article on circumcision can be found here: https://areomagazine.com/2019/09/24/a-wrong-against-boys-an-impossible-conversation-about-circumcision/ Brian Deer, The Doctor Who Fooled the World: Andrew Wakefield’s War on Vaccines (2020) David Robert Grimes, The Irrational Ape: Why Flawed Logic Puts Us All at Risk and How Critical Thinking Can Save the World (2019) Matthew Walker, Why We Sleep: The New Science of Sleep and Dreams (2018) Alexey Guzey, “Matthew Walker's "Why We Sleep" Is Riddled with Scientific and Factual Errors”: https://guzey.com/books/why-we-sleep/ (2019) Andrew Gelman’s scientific blog can be found here: https://statmodeling.stat.columbia.edu/ Timestamps 2:09 Fraudster Paolo Macciarini and his tracheotomies 11:26 Bias towards positive results in scientific publications 15:56 Stuart and his colleagues’ failed attempt to replicate a startling psychology finding 20:03 P-values, p-hacking and the disgrace of Brian Wansink 39:17 Reproducibility failures 42:54 The relationship between power and effect size: Amy Cuddy & power posing; neuroscience; candidate genes; ethical questions in cancer research 01:00:55 Scientists’ own hype when they present their work; the scientific publication system; Andrew Wakefield 01:09:22 How has Stuart managed to avoid becoming disillusioned with all of science & with experts? 01:14:26 How has writing this book affected Stuart’s own practice as a scientist? The Mertonian principles. 01:19:51 How can laypeople do better at critically assessing a scientific finding? 01:24:20 Matthew Walker and his Why We Sleep book.

#27 Modeling the US Presidential Elections, with Andrew Gelman & Merlin Heidemanns

covid-19 stories elections council columbia university presidents stats predict u s american political science review american statistical association andrew gelman

Play Episode Listen Later Nov 1, 2020 60:52

In a few days, a consequential election will take place, as citizens of the United States will go to the polls and elect their president — in fact they already started voting. You probably know a few forecasting models that try to predict what will happen on Election Day — who will get elected, by how much and with which coalition of States? But how do these statistical models work? How do you account for the different sources of uncertainty, be it polling errors, unexpected turnout or media events? How do you model covariation between States? How do you even communicate the model’s results and afterwards assess its performance? To talk about all this, I had the pleasure to talk to Andrew Gelman and Merlin Heidemanns. Andrew was already on episode 20, to talk about his recent book with Jennifer Hill and Aki Vehtari, “Regression and Other Stories”. He’s a professor of statistics and political science at Columbia University and works on a lot of topics, including: why campaign polls are so variable while elections are so predictable, the statistical challenges of estimating small effects, and methods for surveys and experimental design. Merlin is a PhD student in Political Science at Columbia University, and he specializes in political methodology. Prior to his PhD, he did a Bachelor's in Political Science at the Freie Universität Berlin. I hope you’ll enjoy this episode where we dove into the Bayesian model they helped develop for The Economist, and talked more generally about how to forecast elections with statistical methods, and even about the incentives the forecasting industry has as a whole. Thank you to my Patrons for making this episode possible! Visit https://www.patreon.com/learnbayesstats to unlock exclusive Bayesian swag ;) Our theme music is « Good Bayesian », by Baba Brinkman (feat MC Lars and Mega Ran). Check out his awesome work at https://bababrinkman.com/ ! Links from the show: Andrew's website: http://www.stat.columbia.edu/~gelman/ Andrew's blog: https://statmodeling.stat.columbia.edu/ Andrew on Twitter: https://twitter.com/statmodeling Merlin's website: https://merlinheidemanns.github.io/website/ Merlin on Twitter: https://twitter.com/MHeidemanns The Economist POTUS forecast: https://projects.economist.com/us-2020-forecast/president How The Economist presidential forecast works: https://projects.economist.com/us-2020-forecast/president/how-this-works GitHub repo of the Economist model: https://github.com/TheEconomist/us-potus-model Information, incentives, and goals in election forecasts (Gelman, Hullman & Wlezien): http://www.stat.columbia.edu/~gelman/research/unpublished/forecast_incentives3.pdf How to think about extremely unlikely events: https://bit.ly/3ejZYyZ Postal voting could put America’s Democrats at a disadvantage: https://econ.st/3mCxR0P --- Send in a voice message: https://anchor.fm/learn-bayes-stats/message

Can You Still Predict Elections? | Stats + Stories Episode 162

Stats + Stories

Play Episode Listen Later Oct 29, 2020 28:41

With the 2020 U-S presidential election all but upon us, media are rife with prognostications about which way voters are going to swing. Will reliably red states stay red or will voters produce a blue wave that crashes across the country? Will economic uncertainty trump concerns over COVID 19? Is political polarization really as set-in-stone as some have suggested? Understanding voter behavior is a focus of this episode of Stats and Stories where we explore the statistics behind the stories and the stories behind the statistics with guest Andrew Gelman. Andrew Gelman is a professor of statistics and political science at Columbia University. He has received the Outstanding Statistical Application award three times from the American Statistical Association, the award for best article published in the American Political Science Review, and the Council of Presidents of Statistical Societies award for outstanding contributions by a person under the age of 40. His research interests include a wide range of topics, including: why it is rational to vote, why campaign polls are so variable when elections are so predictable and why redistricting is good for democracy among various others.

Statistics is the Least Important Part of Data Science | Andrew Gelman, PhD

The Artists of Data Science

Play Episode Listen Later Oct 12, 2020 57:01

Andrew is an American statistician, professor of statistics and political science, and director of the Applied Statistics Center at Columbia University. He frequently writes about Bayesian statistics, displaying data, and interesting trends in social science. He's also well known for writing posts sharing his thoughts on best statistical practices in the sciences, with a frequent emphasis on what he sees as the absurd and unscientific. FIND ANDREW ONLINE Website: https://statmodeling.stat.columbia.edu/ Twitter: https://twitter.com/StatModeling QUOTES [00:04:16] "We've already passed peak statistics..." [00:05:13] "One thing that we sometimes like to say is that big data need big model because big data are available data. They're not designed experiments, they're not random samples. Often big data means these are measurements. " [00:22:05] "If you design an experiment, you want to know what you're going to do later. So most obviously, you want your sample size to be large enough so that given the effect size that you expect to see, you'll get a strong enough signal that you can make a strong statement." [00:31:00] "The alternative to good philosophy is not no philosophy, it's bad philosophy. " SHOW NOTES [00:03:12] How Dr. Gelman got interested in statistics [00:04:09] How much more hyped has statistical and machine learning become since you first broke into the field? [00:04:44] Where do you see the field of statistical machine learning headed in the next two to five years? [00:06:12] What do you think the biggest positive impact machine learning will have in society in the next two to five years? [00:07:24] What do you think would be some of our biggest concerns in the future? [00:09:07] The thee parts of Bayesian inference [00:12:05] What's the main difference between the frequentist and the Bayesian? [00:13:02] What is a workflow? [00:16:21] Iteratively building models [00:17:50] How does the Bayesian workflow differ from the frequent workflow? [00:18:32] Why is it that what makes this statistical method effective is not what it does with the data, but what data it uses? [00:20:48] Why do Bayesians then tend to be a little bit more skeptical in their thought processes? [00:21:47] Your method of evaluation can be inspired by the model or the model can be inspired by your method of evaluation [00:24:38] What is the usual story when it comes to statistics? And why don't you like it? [00:30:16] Why should statisticians and data scientist care about philosophy? [00:35:04] How can we solve all of our statistics problems using P values? [00:36:14] Is there a difference in interpretations for P-Values between Bayesian and frequentist. [00:36:54] Do you feel like the P value is a difficult concept for a lot of people to understand? And if so, why do you think it's a bit challenging? [00:38:22] Why the least important part of data science is statistics. [00:40:09] Why is it that Americans vote the way they do? [00:42:40] What's the one thing you want people to learn from your story? [00:44:48] The lightning round Special Guest: Andrew Gelman, PhD.

american americans phd columbia university statistics data science bayesian gelman bayesian statistics andrew gelman bayesians harpreet sahota

Philosophy of Data Science | S01 E02 | Scientific Reasoning for Practical Data Science

science philosophy practical communicating scientific data science reasoning andrew gelman

Play Episode Listen Later Sep 30, 2020 55:09

Philosophy of Data Science Series Session 1: Scientific Reasoning for Practical Data Science Episode 2: Scientific Reasoning for Practical Data Science Scientific reasoning plays an essential role in data science and statistics, both for developing new methods and applying our methods to real-world problems. In Session 1's titular episode, Andrew Gelman talks through the role of scientific thinking in his approach to data analysis. He also highlights the good ideas that have been generated by the wider statistical community. Watch it on... YouTube: https://youtu.be/R6mq5Esjzfw Coming up next week: Communicating the Science in Data Science with Kathy Ensor (Rice University & 2022 ASA President) We're always happy to hear your feedback and ideas - just post it in the YouTube comment section to start a conversation. Thank you for your time and support of the series! You can join our mail list at: https://www.podofasclepius.com/mail-list #datascience #statistics #machinelearning #ai #science #stem

Philosophy of Data Science | S01E00 | Welcome to the Series!

Play Episode Listen Later Sep 16, 2020 18:49

The Philosophy of Data Science Series Session 1: Scientific Reasoning for Practical Data Science Episode 0: Welcome to the Philosophy of Data Science Series! This is our very first episode of "The Philosophy of Data Science" series on Pod of Asclepius! We go over our plans for the series plus some thoughts on why data science is such a rich field for discussions on scientific reasoning. Your time is valuable and you deserve a good explanation of why the topics were chosen and how the series is structured to maximize learning. Topic List 0:00 New intro jingle for the series! 0:10 Welcome to the Philosophy of Data Science Series! 1:07 Modes of reasoning 5:33 Session 1 Overview: Scientific Reasoning for Practical Data Science 10:15 Session 2 Overview: Essential Reasoning Skills for Data Science 11:32 Keynotes and Session 4 14:15 Future Sessions Coming up next week: Critical Reasoning in Medical Machine Learning Thank you for your time and support of the series! It only gets better from here! (Seriously, it really does only get better from here. We've got Andrew Gelman coming up, plus Cynthia Rudin, Mihaela van der Schaar...)

philosophy data science modes keynotes mihaela asclepius schaar andrew gelman critical reasoning

#21 Gaussian Processes, Bayesian Neural Nets & SIR Models, with Elizaveta Semenova

Play Episode Listen Later Aug 13, 2020 62:11

I bet you heard a lot about epidemiological compartmental models such as SIR in the last few months? But what are they exactly? And why are they so useful for epidemiological modeling? Elizaveta Semenova will tell you why in this episode, by walking us through the case study she recently wrote with the Stan team. She’ll also tell us how she used Gaussian Processes on spatio-temporal data, to study the spread of Malaria, or to fit dose-response curves in pharmaceutical tests. And finally, she’ll tell us how she used Bayesian neural networks for drug toxicity prediction in her latest paper, and how Bayesian neural nets behave compared to classical neural nets. Ow, and you’ll also learn an interesting link between BNNs and Gaussian Processes… I know: Liza works on _a lot_ of projects! But who is she? Well, she’s a postdoctorate in Bayesian Machine Learning at the pharmaceutical company AstraZeneca, in Cambridge, UK. Elizaveta did her masters in theoretical mathematics in Moscow, Russia, and then worked in financial services as an actuary in various European countries. She then did a PhD in epidemiology at the University of Basel, Switzerland. This is where she got interested in health applications – be it epidemiology, global health or more small-scale biological questions. But she’ll tell you all that in the episode ;) Our theme music is « Good Bayesian », by Baba Brinkman (feat MC Lars and Mega Ran). Check out his awesome work at https://bababrinkman.com/ ! Links from the show: Liza on Twitter: https://twitter.com/liza_p_semenova Liza on GitHub: https://github.com/elizavetasemenova Liza's blog: https://elizavetasemenova.github.io/blog/ A Bayesian neural network for toxicity prediction: https://www.biorxiv.org/content/10.1101/2020.04.28.065532v2 Bayesian Neural Networks for toxicity prediction -- Video presentation: https://www.youtube.com/watch?v=BCQ2oVlu_tY&t=751s Bayesian workflow for disease transmission modeling in Stan: https://mc-stan.org/users/documentation/case-studies/boarding_school_case_study.html Andrew Gelman's comments on the SIR case-study: https://statmodeling.stat.columbia.edu/2020/06/02/this-ones-important-bayesian-workflow-for-disease-transmission-modeling-in-stan/ Determining organ weight toxicity with Bayesian causal models: https://www.biorxiv.org/content/10.1101/754853v1 Material for Applied Machine Learning Days ("Embracing uncertainty"): https://github.com/elizavetasemenova/EmbracingUncertainty Predicting Drug-Induced Liver Injury with Bayesian Machine Learning: https://pubs.acs.org/doi/abs/10.1021/acs.chemrestox.9b00264 Ordered Logistic Regression in Stan, PyMC3 and Turing: https://medium.com/@liza_p_semenova/ordered-logistic-regression-and-probabilistic-programming-502d8235ad3f PyMCon website: https://pymc-devs.github.io/pymcon/ PyMCon Call For Proposal: https://pymc-devs.github.io/pymcon/cfp PyMCon Sponsorship Form: https://docs.google.com/forms/d/e/1FAIpQLSdRDI1z0U0ZztONOFiZt2VdsBIZtAWB4JAUA415Iw8RYqNbXQ/viewform PyMCon Volunteer Form: https://docs.google.com/forms/d/e/1FAIpQLScCLW5RkNtBz1u376xwelSsNpyWImFisSMjZGP35fYi2QHHXw/viewform --- Send in a voice message: https://anchor.fm/learn-bayes-stats/message

#20 Regression and Other Stories, with Andrew Gelman, Jennifer Hill & Aki Vehtari

Play Episode Listen Later Jul 30, 2020 63:44

Once upon a time, there was an enchanted book filled with hundreds of little plots, applied examples and linear regressions — the prettiest creature that was ever seen. Its authors were excessively fond of it, and its readers loved it even more. This magical book had a nice blue cover made for it, and everybody aptly called it « Regression and other Stories »! As every good fairy tale, this one had its share of villains — the traps where statistical methods fall and fail you; the terrible confounders, lurking in the dark; the ill-measured data that haunt your inferences! But once you defeat these monsters, you’ll be able to think about, build and interpret regression models. This episode will be filled with stories — stories about linear regressions! Here to narrate these marvelous statistical adventures are Andrew Gelman, Jennifer Hill and Aki Vehtari — the authors of the brand new Regression and other Stories. Andrew is a professor of statistics and political science at Columbia University. Jennifer is a professor of applied statistics at NYU. She develops methods to answer causal questions related to policy research and scientific development. Aki is an associate professor in computational probabilistic modeling at Aalto University, Finland. In this episode, they tell us why they wrote this book, who it is for and they also give us their 10 tips to improve your regression modeling! We also talked about the limits of regression and about going to Mars… Other good news: until October 31st 2020, you can go to http://www.cambridge.org/wm-ecommerce-web/academic/landingPage/GoodBayesian2020 and buy the book with a 20% discount by entering the promo code “GoodBayesian2020” upon checkout! That way, you’ll make up your own stories before going to sleep and dream of a world where we can easily generalize from sample to population, and where multilevel regression with poststratification is a bliss… Our theme music is « Good Bayesian », by Baba Brinkman (feat MC Lars and Mega Ran). Check out his awesome work at https://bababrinkman.com/ ! Links from the show: Regression and Other Stories on Cambridge Press website: http://www.cambridge.org/wm-ecommerce-web/academic/landingPage/GoodBayesian2020 Amazon page (because of VAT laws, in some regions ordering from Amazon can be cheaper than from the editor directly, even with the discount): https://www.amazon.com/Regression-Stories-Analytical-Methods-Research/dp/110702398X Code, data and examples for the book: https://avehtari.github.io/ROS-Examples/ Port of the book in Python and Bambi: https://github.com/bambinos/Bambi_resources/tree/master/ROS Andrew's home page: http://www.stat.columbia.edu/~gelman/ Andrew's blog: https://statmodeling.stat.columbia.edu/ Andrew on Twitter: https://twitter.com/statmodeling Jennifer's home page: https://steinhardt.nyu.edu/people/jennifer-hill Aki's teaching material: https://avehtari.github.io/ Aki's home page: https://users.aalto.fi/~ave/ Aki on Twitter: https://twitter.com/avehtari --- Send in a voice message: https://anchor.fm/learn-bayes-stats/message

amazon stories finland columbia university nyu python regression vat other stories aki mega ran aalto university mc lars jennifer hill baba brinkman andrew gelman cambridge press

#SpecialAnnouncement: Patreon Launched!

previous slack launched btw regression turing mega ran bayes mc lars jennifer hill baba brinkman andrew gelman bayesians

Play Episode Listen Later Jun 26, 2020 7:38

I hope you’re all safe! Some of you also asked me if I had set up a Patreon so that they could help support the show, and that’s why I’m sending this short special episode your way today. I had thought about that, but I wasn’t sure there was a demand for this. Apparently, there is one — at least a small one — so, first, I wanna thank you and say how grateful I am to be in a community that values this kind of work! The Patreon page is now live at patreon.com/learnbayesstats. It starts as low as 3€ and you can pick from 4 different tiers: "Maximum A Posteriori" (3€): Join the Slack, where you can ask questions about the show, discuss with like-minded Bayesians and meet them in-person when you travel the world. "Full Posterior" (5€): Previous tier + Your name in all the show notes, and I'll express my gratitude to you in the first episode to go out after your contribution. You also get early access to the special episodes. -- that I'll make at an irregular pace and will include panel discussions, book releases, live shows, etc. "Principled Bayesian" (20€): Previous tiers + Every 2 months, I'll ask my guest two questions voted-on by "Principled Bayesians". I'll probably do that with a poll in the Slack channel, which will be only answered by the "Principled Bayesians" and of these questions, I will ask the top 2 every two months on the show. "Good Bayesian" (200€, only 8 spots): Previous tiers + Every 2 months, you can come on the show and you ask one question to the guest without a vote. So that's why I can't have too many people in that tier. Before telling you the best part: I already have a lot of ideas for exclusive content and options. I first need to see whether you're as excited as I am about it. If I see you are, I'll be able to add new perks to the tiers! So give me your feedback about the current tiers or any benefits you'd like to see there... but don't see yet! BTW, you have a new way to do that now: sending me voice messages at anchor.fm/learn-bayes-stats/message! Now, the icing on the cake: until July 31st, if you choose the "Full Posterior" tier (5$) or higher, you get early access to the very special episode I'm planning with Andrew Gelman, Jennifer Hill and Aki Vehtari about their upcoming book, "Regression and other stories". To top it off, there will be a promo code in the episode to buy the book at a discount price — now, that is an offer you can't turn down! Alright, that is it for today — I hope you’re as excited as I am for this new stage in the podcast’s life! Please keep the emails, the tweets, the voice messages, the carrier pigeons coming with your feedback, questions and suggestions. In the meantime, take care and I’ll see you in the next episode — episode 19, with Cameron Pfiffer, who’s the first economist to come on the show and who’s a core-developer of Turing.jl. We’re gonna talk about the Julia probabilistic programming landscape, Bayes in economics and causality — it’s gonna be fun ;) Again, patreon.com/learnbayesstats if you want to support the show and unlock some nice perks. Thanks again, I am very grateful for any support you can bring me! Our theme music is « Good Bayesian », by Baba Brinkman (feat MC Lars and Mega Ran). Check out his awesome work at https://bababrinkman.com/ ! Links from the show: LBS Patreon page: patreon.com/learnbayesstats Send me voice messages: anchor.fm/learn-bayes-stats/message --- Send in a voice message: https://anchor.fm/learn-bayes-stats/message

Ep. 81 - Junk Science, Click Bait, and Meaningless Experiments

The Local Maximum

Play Episode Listen Later Aug 27, 2019 49:24

Ever feel like there's something wrong with all the news and information you consume, even when you get it from professional scientists and doctors? We tackle the reproducibility crisis in science today starting with taking a look at some mudslinging between high profile online academics Andrew Gelman and Cass Sunstein. We also talk about the probable over-publishing of studies in stock market prediction and elsewhere, and scientists reading too much when interpreting results to fit their narrative and ideology. localmaxradio.com/81

experiments clickbait meaningless cass sunstein junk science andrew gelman

Significant Science ft. Dr Andrew Gelman

The Neoliberal Podcast

Play Episode Listen Later Jan 19, 2019 53:37

Jeremiah sits down with Dr. Andrew Gelman to discuss the replication crisis in science - how it began, what we know and how we should think about science moving forward. Patreon subscribers get access to bonus episodes and sticker of the month club. If you like what we do (and want stickers each month!) please consider supporting us at patreon.com/neoliberalproject.

science significant andrew gelman

#43 Election Forecasting and Polling

DataFramed

Play Episode Listen Later Oct 7, 2018 65:14 Transcription Available

Hugo speaks with Andrew Gelman about statistics, data science, polling, and election forecasting. Andy is a professor of statistics and political science and director of the Applied Statistics Center at Columbia University and this week we’ll be talking the ins and outs of general polling and election forecasting, the biggest challenges in gauging public opinion, the ever-present challenge of getting representative samples in order to model the world and the types of corrections statisticians can and do perform. "Chatting with Andy was an absolute delight and I cannot wait to share it with you!"-Hugo Links from the show FROM THE INTERVIEWAndrew's Blog Andrew on Twitter We Need to Move Beyond Election-Focused Polling (Gelman and Rothschild, Slate)We Gave Four Good Pollsters the Same Raw Data. They Had Four Different Results (Cohn, The New York Times).19 things we learned from the 2016 election (Gelman and Azari, Science, 2017)The best books on How Americans Vote (Gelman, Five Books)The best books on Statistics (Gelman, Five Books)Andrew's Research FROM THE SEGMENTSStatistical Lesson of the Week (with Emily Robinson at ~13:30)The five Cs (Loukides, Mason, and Patil, O'Reilly)Data Science Best Practices (with Ben Skrainka~40:40)Oberkampf & Roy’s Verification and Validation in Scientific Computing provides a thorough yet very readable treatment A comprehensive framework for verification, validation, and uncertainty quantification in scientific computing (Roy and Oberkampf, Science Direct) Original music and sounds by The Sticks.

Ep. 56: Straight Outta Chumptown

NonProphets

Play Episode Listen Later Apr 25, 2018

Episode 56 of the NonProphets podcast, in which Atief, Robert, and Scott discuss "the backfire effect"—the fact that people sometimes seem to hold ideas even more firmly after being confronted with evidence they are wrong. We talk about Andrew Gelman's skepticism of the recent New England Journal of Medicine finding that firearm injuries in the US drop 20% while NRA members are attending national meetings (00:50); how skeptical we should be of research that seem to confirm our preconceptions (04:14); whether evidence can change people's minds (15:27); "The Debunking Handbook" techniques for correcting false ideas (18:48); research into how much contrary information it takes to change people's minds (23:21); whether culture or economics determines election results (29:11); how we avoid bias in forecasting and decision-making (31:25); and how we can stay out of Chumptown (39:20). As always, you can reach us at nonprophetspod.wordpress.com or at nonprophetspod@gmail.com. (recorded 4/4/2018)

medicine nra new england journal straight outta andrew gelman nonprophets

Andrew Gelman on Social Science, Small Samples, and the Garden of the Forking Paths

EconTalk

Play Episode Listen Later Mar 20, 2017 67:42

Statistician, blogger, and author Andrew Gelman of Columbia University talks with EconTalk host Russ Roberts about the challenges facing psychologists and economists when using small samples. On the surface, finding statistically significant results in a small sample would seem to be extremely impressive and would make one even more confident that a larger sample would find even stronger evidence. Yet, larger samples often fail to lead to replication. Gelman discusses how this phenomenon is rooted in the incentives built into human nature and the publication process. The conversation closes with a general discussion of the nature of empirical work in the social sciences.

psychology gardens economics columbia university statistics social sciences samples for king statisticians gelman russ roberts econtalk forking paths statistical significance andrew gelman econlib

Ep. 128: Andrew Gelman Clarifies Some Misconceptions About the 2016 Election, Mama DeVega Returns, and Jane Elliott Chimes In

The Chauncey DeVega Show

Play Episode Listen Later Mar 17, 2017 78:44

This week's episode of The Chauncey DeVega Show features three great guests. Political scientist and statistician Andrew Gelman is the first guest. He is a professor at Columbia University and the author of several books including Red State, Blue State, Rich State, Poor State: Why Americans Vote the Way They Do. He recently wrote a great piece for Slate called "19 Lessons for Political Scientists from the 2016 Election". During this week's episode, Dr. Gelman does some great teaching and sharing about the 2016 presidential election and what the so-called "smart people" got right and wrong. Civil rights activist and educator Jane Elliott chimed in for the second segment of this week's podcast. She is most famous for her much discussed and documented Blue Eyes/Brown Eyes teaching and learning exercise from the 1960s. In the five decades since, Sister Elliot has not stopped speaking speak truth to power about the color line, prejudice, and bigotry in America and around the world. This is a preview of next week's full episode with Jane Elliott. Mama DeVega also makes her return to The Chauncy DeVega Show. She gives thanks for the kind birthday gifts that the friends of the podcast and Chauncey DeVega sent her way. Mama DeVega and Chauncey also debate the merits of attending one's own funeral while still alive and how best to fake a death. Mama DeVega also shares a story about her encounter--or so she believes--with one of the 9-11 hijackers. During this week's podcast, Chauncey talks about Sean Spicer's press conference where he defended Donald Trump's lies about wiretapping, the evil Trump 2018 budget and its plans to kill the "useless eaters", and gives out a "you big dummy award" to Rachel Maddow.

america donald trump lessons elections political mama columbia university misconceptions civil slate red state rachel maddow chimes sean spicer blue states clarifies 2016 election chauncey political scientists gelman jane elliott andrew gelman way they do rich state chauncey devega chauncey devega show

Hackademics II: The Hackers

Hi-Phi Nation

Play Episode Listen Later Mar 14, 2017 44:25

One scientist decided to put the entire field of psychology to test to see how many of its findings hold up to scrutiny. At the same time, he had scientists bet on the success-rate of their own field. We look at the surprising paradoxes of humans being human, trying to learn about humans, and the elusive knowledge of human nature. Guest voices include Brian Nosek of the Center for Open Science, Andrew Gelman of Columbia University, Deborah Mayo of Virginia Tech, and Matthew Makel of Duke TiP. A philosophical take on the replication crisis in the sciences. Learn more about your ad choices. Visit megaphone.fm/adchoices

columbia university hackers virginia tech open science brian nosek andrew gelman

Episode 99: Mockingbirds, Destructo-Critics, and Mr. Robot

Very Bad Wizards

Play Episode Listen Later Sep 27, 2016 123:28

David and Tamler tackle three topics on their last double digit episode. First, should a middle school perform "To Kill a Mockingbird" even if they have to use bad language the "n-word," and talk about sexual assault? Tamler relates a story involving his daughter (who was supposed to play Scout) and a playwright who refused to allow his play to be censored. But when it comes to drama, middle school's got nothing on social psychology. Next, David and Tamler break down the latest controversy surrounding Princeton psychologist Susan Fiske's leaked column about the bullying destructo-critics and methodological terrorists that are challenging the establishment in the field. Finally, they give a spoiler-filled analysis of season 2 of Mr. Robot, a polarizing season for many fans. Tamler's suffering from a little theory fatigue, but David blows his mind with his explanation of what's really going on with the Dark Army and F-Society. Have you ever cried during sex?LinksTo Kill a Mockingbird stage play [stageagent.com]Mob Rule or the Wisdom of Crowds? Susan Fiske's forthcoming column in the APS Observer [verybadwizards.com]Andrew Gelman's blog post about Susan Fiske's column [andrewgelman.com]Ioannidis, J. P. (2005). Why most published research findings are false. PLoS Med, 2(8), e124. [plos.org]The Hardest Science blog by Sanjay Srivastava (@hardsci)sometimes i'm wrong blog by Simine Vazire (@siminevazire)The 20% Statistician blog by Daniel Lakens (@lakens)Too Many Cooks [youtube.com]Bitcoin explained and made simple [youtube.com]Key generation [wikipedia.org]

wisdom robots bitcoin critics crowds mockingbird to kill statisticians fiske mob rule destructo too many cooks andrew gelman dark army sanjay srivastava plos med tamler simine vazire

Rationally Speaking #147 - Andrew Gelman on "Why do Americans vote the way they do?"

Rationally Speaking

Play Episode Listen Later Nov 15, 2015 52:42

america americans vote republicans democrats andrew gelman rationally speaking

Advantage and Disadvantage (Bonus Action 001)

Play Episode Listen Later Dec 10, 2014 14:32

Welcome to the inaugural episode of the Bonus Action podcast. This show is about the rules of the 5th edition of Dungeons and Dragons and we plan to explore those rules one by one. In this episode Sam and James discuss the Advantage/Disadvantage mechanic. You can find an explanation of this rule in the Basic D&D PDF on page 57 or in the 5e D&D Player's Handbook on page 173. Links: D&D 5e Basic PDF: http://dnd.wizards.com/articles/features/basicrules James' blog: http://www.worldbuilderblog.me Sam's blog: http://www.rpgmusings.com Eric Michaels Music: http://www.theericmichaels.com/ Probabilities and +/- bonus discussions: Online Dungeon Master: http://onlinedungeonmaster.com/2012/05/24/advantage-and-disadvantage-in-dd-next-the-math/ Andrew Gelman: http://andrewgelman.com/2014/07/12/dnd-5e-advantage-disadvantage-probability/ Zero Hit Points: http://www.zerohitpoints.com/Advantage-in-DnD-5 Critical Hits: http://www.critical-hits.com/blog/2012/06/11/dd-advantage-vs-flat-bonuses/ Probability Comparison Chart: http://worldbuilderblog.files.wordpress.com/2014/02/screen-shot-2014-02-21-at-8-09-06-pm.png Old list of advantage and disadvantage situations: http://www.reddit.com/r/DnD/comments/29xk0j/5e_summary_of_situations_causing_advantage_or/ Support the show, shop below...

action dungeons and dragons advantage handbook probability disadvantage andrew gelman bonus action

Advantage and Disadvantage (Bonus Action 001)

action dragons dungeons and dragons advantage dungeons handbook probability disadvantage andrew gelman

Play Episode Listen Later Dec 10, 2014 14:32

Welcome to the inaugural episode of the Bonus Action podcast. This show is about the rules of the 5th edition of Dungeons and Dragons and we plan to explore those rules one by one. In this episode Sam and James discuss the Advantage/Disadvantage mechanic. You can find an explanation of this rule in the Basic D&D PDF on page 57 or in the 5e D&D Player’s Handbook on page 173.Links:D&D 5e Basic PDF: http://dnd.wizards.com/articles/features/basicrulesJames’ blog: http://www.worldbuilderblog.meSam’s blog: http://www.rpgmusings.comEric Michaels Music: http://www.theericmichaels.com/Probabilities and +/- bonus discussions:Online Dungeon Master: http://onlinedungeonmaster.com/2012/05/24/advantage-and-disadvantage-in-dd-next-the-math/Andrew Gelman: http://andrewgelman.com/2014/07/12/dnd-5e-advantage-disadvantage-probability/Zero Hit Points: http://www.zerohitpoints.com/Advantage-in-DnD-5Critical Hits: http://www.critical-hits.com/blog/2012/06/11/dd-advantage-vs-flat-bonuses/Probability Comparison Chart: http://worldbuilderblog.files.wordpress.com/2014/02/screen-shot-2014-02-21-at-8-09-06-pm.pngOld list of advantage and disadvantage situations: http://www.reddit.com/r/DnD/comments/29xk0j/5e_summary_of_situations_causing_advantage_or/Support the show, shop below...

Advantage and Disadvantage (Bonus Action 001)

action dragons dungeons and dragons advantage dungeons handbook probability disadvantage andrew gelman

Play Episode Listen Later Dec 10, 2014 14:32

Welcome to the inaugural episode of the Bonus Action podcast. This show is about the rules of the 5th edition of Dungeons and Dragons and we plan to explore those rules one by one. In this episode Sam and James discuss the Advantage/Disadvantage mechanic. You can find an explanation of this rule in the Basic D&D PDF on page 57 or in the 5e D&D Player’s Handbook on page 173.Links:D&D 5e Basic PDF: http://dnd.wizards.com/articles/features/basicrulesJames’ blog: http://www.worldbuilderblog.meSam’s blog: http://www.rpgmusings.comEric Michaels Music: http://www.theericmichaels.com/Probabilities and +/- bonus discussions:Online Dungeon Master: http://onlinedungeonmaster.com/2012/05/24/advantage-and-disadvantage-in-dd-next-the-math/Andrew Gelman: http://andrewgelman.com/2014/07/12/dnd-5e-advantage-disadvantage-probability/Zero Hit Points: http://www.zerohitpoints.com/Advantage-in-DnD-5Critical Hits: http://www.critical-hits.com/blog/2012/06/11/dd-advantage-vs-flat-bonuses/Probability Comparison Chart: http://worldbuilderblog.files.wordpress.com/2014/02/screen-shot-2014-02-21-at-8-09-06-pm.pngOld list of advantage and disadvantage situations: http://www.reddit.com/r/DnD/comments/29xk0j/5e_summary_of_situations_causing_advantage_or/Support the show, shop below...

Advantage and Disadvantage (Bonus Action 001)