Podcast appearances and mentions of monte carlo mcmc

Sparse Bayesian mass-mapping using trans-dimensional MCMC by Augustin Marignier et al. on Monday 28 November Uncertainty quantification is a crucial step of cosmological mass-mapping that is often ignored. Suggested methods are typically only approximate or make strong assumptions of Gaussianity of the shear field. Probabilistic sampling methods, such as Markov chain Monte Carlo (MCMC), draw samples form a probability distribution, allowing for full and flexible uncertainty quantification, however these methods are notoriously slow and struggle in the high-dimensional parameter spaces of imaging problems. In this work we use, for the first time, a trans-dimensional MCMC sampler for mass-mapping, promoting sparsity in a wavelet basis. This sampler gradually grows the parameter space as required by the data, exploiting the extremely sparse nature of mass maps in wavelet space. The wavelet coefficients are arranged in a tree-like structure, which adds finer scale detail as the parameter space grows. We demonstrate the trans-dimensional sampler on galaxy cluster-scale images where the planar modelling approximation is valid. In high-resolution experiments, this method produces naturally parsimonious solutions, requiring less than 1% of the potential maximum number of wavelet coefficients and still producing a good fit to the observed data. In the presence of noisy data, trans-dimensional MCMC produces a better reconstruction of mass-maps than the standard smoothed Kaiser-Squires method, with the addition that uncertainties are fully quantified. This opens up the possibility for new mass maps and inferences about the nature of dark matter using the new high-resolution data from upcoming weak lensing surveys such as Euclid. arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2211.13963v1

Sparse Bayesian mass-mapping using trans-dimensional MCMC by Augustin Marignier et al. on Sunday 27 November Uncertainty quantification is a crucial step of cosmological mass-mapping that is often ignored. Suggested methods are typically only approximate or make strong assumptions of Gaussianity of the shear field. Probabilistic sampling methods, such as Markov chain Monte Carlo (MCMC), draw samples form a probability distribution, allowing for full and flexible uncertainty quantification, however these methods are notoriously slow and struggle in the high-dimensional parameter spaces of imaging problems. In this work we use, for the first time, a trans-dimensional MCMC sampler for mass-mapping, promoting sparsity in a wavelet basis. This sampler gradually grows the parameter space as required by the data, exploiting the extremely sparse nature of mass maps in wavelet space. The wavelet coefficients are arranged in a tree-like structure, which adds finer scale detail as the parameter space grows. We demonstrate the trans-dimensional sampler on galaxy cluster-scale images where the planar modelling approximation is valid. In high-resolution experiments, this method produces naturally parsimonious solutions, requiring less than 1% of the potential maximum number of wavelet coefficients and still producing a good fit to the observed data. In the presence of noisy data, trans-dimensional MCMC produces a better reconstruction of mass-maps than the standard smoothed Kaiser-Squires method, with the addition that uncertainties are fully quantified. This opens up the possibility for new mass maps and inferences about the nature of dark matter using the new high-resolution data from upcoming weak lensing surveys such as Euclid. arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2211.13963v1

Python 3.11 und Listen 15. November 2022, Jochen Johannes, Dominik und Jochen unterhalten sich über Python 3.11. Hauptthema hätte eigentlich Listen als Datenstruktur sein sollen, aber zu Python 3.11 gab es dann doch etwas mehr zu sagen, daher war das dann nicht so ausführlich wie geplant. Überhaupt hatten wir diesmal recht viele Abschweifungen und Nebenthemen drin. Aber gut, mit den Kapitelmarken sollte man die auch skippen können. Vielleicht dauert es ja auch nicht mehr so lang bis zur nächsten Episode

Strong Gravitational Lensing Parameter Estimation with Vision Transformer by Kuan-Wei Huang et al. on Monday 10 October Quantifying the parameters and corresponding uncertainties of hundreds of strongly lensed quasar systems holds the key to resolving one of the most important scientific questions: the Hubble constant ($H_{0}$) tension. The commonly used Markov chain Monte Carlo (MCMC) method has been too time-consuming to achieve this goal, yet recent work has shown that convolution neural networks (CNNs) can be an alternative with seven orders of magnitude improvement in speed. With 31,200 simulated strongly lensed quasar images, we explore the usage of Vision Transformer (ViT) for simulated strong gravitational lensing for the first time. We show that ViT could reach competitive results compared with CNNs, and is specifically good at some lensing parameters, including the most important mass-related parameters such as the center of lens $theta_{1}$ and $theta_{2}$, the ellipticities $e_1$ and $e_2$, and the radial power-law slope $gamma'$. With this promising preliminary result, we believe the ViT (or attention-based) network architecture can be an important tool for strong lensing science for the next generation of surveys. The open source of our code and data is in url{https://github.com/kuanweih/strong_lensing_vit_resnet}. arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2210.04143v1

Understanding if molecular ratios can be used as diagnostics of AGN and starburst activity: The case of NGC 1068 by J. Butterworth et al. on Thursday 22 September Molecular line ratios, such as HCN(1-0)/HCO$^+$(1-0) and HCN(4-3)/CS(7-6), are routinely used to identify active galactic nuclei (AGN) activity in galaxies. Such ratios are, however, hard to interpret as they are highly dependent on the physics and energetics of the gas, and hence can seldom be used as a unique, unambiguous diagnostic. We used the composite galaxy NGC 1068 as a `laboratory' to investigate whether molecular line ratios between HCN, HCO$^+$, and CS are useful tracers of AGN-dominated gas and determine the origin of the differences in such ratios across different types of gas. Such a determination will enable a more rigorous use of such ratios. First, we empirically examined the aforementioned ratios at different angular resolutions to quantify correlations. We then used local thermodynamic equilibrium (LTE) and non-LTE analyses coupled with Markov chain Monte Carlo (MCMC) sampling in order to determine the origin of the underlying differences in ratios. We propose that at high spatial resolution (< 50 pc) the HCN(4-3)/CS(2-1) is a reliable tracer of AGN activity. We also find that the variations in ratios are not a consequence of different densities or temperature but of different fractional abundances, yielding to the important result that it is essential to consider the chemical processes at play when drawing conclusions from radiative transfer calculations. From analyses at varying spatial scales, we find that previously proposed molecular line ratios, as well as a new one, have varying levels of consistency. We also determine from an investigation of radiative transfer modelling of our data that it is essential to consider the chemistry of the species when reaching conclusions from radiative transfer calculations. arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2209.05928v2

Understanding if molecular ratios can be used as diagnostics of AGN and starburst activity: The case of NGC 1068 by J. Butterworth et al. on Thursday 22 September Molecular line ratios, such as HCN(1-0)/HCO$^+$(1-0) and HCN(4-3)/CS(7-6), are routinely used to identify active galactic nuclei (AGN) activity in galaxies. Such ratios are, however, hard to interpret as they are highly dependent on the physics and energetics of the gas, and hence can seldom be used as a unique, unambiguous diagnostic. We used the composite galaxy NGC 1068 as a `laboratory' to investigate whether molecular line ratios between HCN, HCO$^+$, and CS are useful tracers of AGN-dominated gas and determine the origin of the differences in such ratios across different types of gas. Such a determination will enable a more rigorous use of such ratios. First, we empirically examined the aforementioned ratios at different angular resolutions to quantify correlations. We then used local thermodynamic equilibrium (LTE) and non-LTE analyses coupled with Markov chain Monte Carlo (MCMC) sampling in order to determine the origin of the underlying differences in ratios. We propose that at high spatial resolution (< 50 pc) the HCN(4-3)/CS(2-1) is a reliable tracer of AGN activity. We also find that the variations in ratios are not a consequence of different densities or temperature but of different fractional abundances, yielding to the important result that it is essential to consider the chemical processes at play when drawing conclusions from radiative transfer calculations. From analyses at varying spatial scales, we find that previously proposed molecular line ratios, as well as a new one, have varying levels of consistency. We also determine from an investigation of radiative transfer modelling of our data that it is essential to consider the chemistry of the species when reaching conclusions from radiative transfer calculations. arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2209.05928v2

A self-consistent dynamical model of the Milky Way disc adjusted to Gaia data by A. C. Robin et al. on Tuesday 13 September This paper shows how a self-consistent dynamical model can be obtained by fitting the gravitational potential of the Milky Way to the stellar kinematics and densities from Gaia data. Using the Besancon Galaxy Model we derive a potential and the disc stellar distribution functions are computed based on three integrals of motion to model stationary stellar discs. The gravitational potential and the stellar distribution functions are built self-consistently, and then adjusted to be in agreement with the kinematics and the density distributions obtained from Gaia observations. A Markov chain Monte Carlo (MCMC) is used to fit the free parameters of the dynamical model to Gaia parallax and proper motion distributions. The fit is done on several sets of Gaia eDR3 data, widely spread in longitudes and latitudes. We are able to determine the velocity dispersion ellipsoid and its tilt for sub-components of different ages, both varying with R and z. The density laws and their radial scale lengths, for the thin and thick disc populations are also obtained self-consistently. This new model has some interesting characteristics, such as a flaring thin disc. The thick disc is found to present very distinctive characteristics from the old thin disc, both in density and kinematics. This well supports the idea that thin and thick discs were formed in distinct scenarios as the density and kinematics transition between them is found to be abrupt. The dark matter halo is shown to be nearly spherical. We also derive the Solar motion to be (10.79 $pm$ 0.56, 11.06 $pm$ 0.94, 7.66 $pm$ 0.43) km/s, in good agreement with recent studies. The resulting fully self-consistent gravitational potential, still axisymmetric, is a good approximation of a smooth mass distribution in the Milky Way and can be used for further studies, including to compute orbits for real stars in our Galaxy (abridged). arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2208.13827v2

Variable Chaplygin Gas: Constraints from Supernovae and Gravitational Wave Merger Events by Ashley Chraya et al. on Monday 12 September We investigate the cosmological constraints on the variable Chaplygin gas model from the latest observational data: SCP Union 2.1 compilation dataset of Type Ia supernovae (SNe Ia), Pantheon sample of SNe Ia and GWTC-3 of gravitational wave merger events. The variable Chaplygin gas is a model of interacting dark matter and dark energy which interpolates from dust-dominated era to quintessence dominated era. The variable Chaplygin gas model is shown to be compatible with Type Ia Supernovae and gravitational merger data. We have obtained tighter constraints on cosmological parameters $Omega_m$ and $n$, using the Pantheon sample. By using the Markov chain Monte Carlo (MCMC) method on the Pantheon sample, we obtain $Omega_m$=0.108 $pm$ 0.034, n=1.157 $pm$ 0.513 and $H_0$=70.020 $pm$ 0.407 and on GWTC-3, we obtain $Omega_m$=0.130 $pm$ 0.076, n=0.897 $pm$ 1.182 and $H_0$=69.838 $pm$ 3.007. arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2206.14192v2

5.02 picks at some TGFBI picks (0:40), self-identifies favorite pick/worst pick so far (1:49), poses a new book club target: Beren and Lúthien (8:46), wanders through global ADP trends in TGFBI (11:39), makes a Markov chain Monte Carlo (MCMC) model of the TGFBI draft that answers the question: when will this guy be available (14:43), and reviews Trevor Story (19:35).

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.10.10.334003v1?rss=1 Authors: Redelings, B. D. Abstract: BAli-Phy is a Markov chain Monte Carlo (MCMC) program that jointly estimates phylogeny, alignment, and other parameters from unaligned sequence data. Version 3 is substantially faster for large trees, and implements covarion models, RNA stem models, and other new models. It implements ancestral state reconstruction, allows prior selection for all model parameters, and can also analyze multiple genes simultaneously. Availability: Software is available for download at http://www.bali-phy.org. C++ source code is freely available on Github under the GPL2 License. Copy rights belong to original authors. Visit the link for more info

Show Notes: (2:02) Peadar discussed his undergraduate experience studying Physics and Philosophy at the University of Bristol. (3:05) Peadar then pursued a Master’s degree in Mathematics from the University of Luxembourg, where he did a thesis on machine learning for time series forecasting. (4:16) Peadar commented on his varied work experience with various companies, particularly on data maturity and the difference of established companies and startups. (7:11) Peadar talked about his latest startup called aflorithmic Labs, which develops tech platform that powers and enables the creation of a new generation hyper-personalized / super-relevant podcasts. (8:13) In the series “Interviews with Data Scientists,” Peadar interviewed with 24 of the world’s most influential and innovative data scientists from across the spectrum. He talked about the common traits in the best data scientists. (10:05) Peadar mentioned his contribution to PyMC3, a Python package for Bayesian statistical modeling and Probabilistic Machine Learning focusing on advanced Markov chain Monte Carlo (MCMC) and variational inference (VI) algorithms. (11:32) Peadar talked about the probabilistic programming survey he conducted recently, in which A/B testing is a big use case. (13:37) In his talk “Lies damned lies and statistics in Python” at PyData London 2016, Peadar compared and debugged models in Statsmodels, scikit-learn and PyMC3. He recalled the differences here. (15:27) Peadar went over “Probabilistic Programming Primer” - an online course he designed to teach people to learn how to enhance modeling abilities and better communicate risk. (18:32) Peadar talked about the recent development in the PyData ecosystem, in reference to his talk “A Map of the PyData Stack” at PyData Amsterdam 2016. (20:18) Discussing his blog post “How to successfully deliver Data Science in the Enterprise,” Peadar went over the people, processes, and things that are required to make data science a successful component in enterprise businesses. (23:25) Discussing his blog post “Building Full-Stack Vertical Data Products,” Peadar emphasized the importance of providing end-to-end value with lean metrics as a data scientist. (29:50) Discussing his blog post “One weird tips to improve the success of DS projects,” Peadar shared his small practice of writing down the risks before embarking on a project. (32:58) Discussing his blog post “3 pitfalls for non-technical managers managing DS teams,” Peadar described the things that non-technical managers will get wrong in managing a technical project. (35:31) Discussing his blog post “What does it mean to be a Senior DS?,” Peadar explained why senior data scientists should understand the soft side of technical decision making and should care about ethics. (38:57) Peadar gave a brief overview of machine learning interpretability. (40:21) Closing segments. His Contact Info: LinkedIn Twitter GitHub Medium Quora Website His Recommended Resources: LIME SHAP Stitch Fix Tech Blog Ravelin Blog Stripe Engineering Blog Spotify Discover Weekly Dale Carnegie’s How to Win Friends and Influence People

Show Notes: (2:02) Peadar discussed his undergraduate experience studying Physics and Philosophy at the University of Bristol. (3:05) Peadar then pursued a Master’s degree in Mathematics from the University of Luxembourg, where he did a thesis on machine learning for time series forecasting. (4:16) Peadar commented on his varied work experience with various companies, particularly on data maturity and the difference of established companies and startups. (7:11) Peadar talked about his latest startup called aflorithmic Labs, which develops tech platform that powers and enables the creation of a new generation hyper-personalized / super-relevant podcasts. (8:13) In the series “Interviews with Data Scientists,” Peadar interviewed with 24 of the world’s most influential and innovative data scientists from across the spectrum. He talked about the common traits in the best data scientists. (10:05) Peadar mentioned his contribution to PyMC3, a Python package for Bayesian statistical modeling and Probabilistic Machine Learning focusing on advanced Markov chain Monte Carlo (MCMC) and variational inference (VI) algorithms. (11:32) Peadar talked about the probabilistic programming survey he conducted recently, in which A/B testing is a big use case. (13:37) In his talk “Lies damned lies and statistics in Python” at PyData London 2016, Peadar compared and debugged models in Statsmodels, scikit-learn and PyMC3. He recalled the differences here. (15:27) Peadar went over “Probabilistic Programming Primer” - an online course he designed to teach people to learn how to enhance modeling abilities and better communicate risk. (18:32) Peadar talked about the recent development in the PyData ecosystem, in reference to his talk “A Map of the PyData Stack” at PyData Amsterdam 2016. (20:18) Discussing his blog post “How to successfully deliver Data Science in the Enterprise,” Peadar went over the people, processes, and things that are required to make data science a successful component in enterprise businesses. (23:25) Discussing his blog post “Building Full-Stack Vertical Data Products,” Peadar emphasized the importance of providing end-to-end value with lean metrics as a data scientist. (29:50) Discussing his blog post “One weird tips to improve the success of DS projects,” Peadar shared his small practice of writing down the risks before embarking on a project. (32:58) Discussing his blog post “3 pitfalls for non-technical managers managing DS teams,” Peadar described the things that non-technical managers will get wrong in managing a technical project. (35:31) Discussing his blog post “What does it mean to be a Senior DS?,” Peadar explained why senior data scientists should understand the soft side of technical decision making and should care about ethics. (38:57) Peadar gave a brief overview of machine learning interpretability. (40:21) Closing segments. His Contact Info: LinkedIn Twitter GitHub Medium Quora Website His Recommended Resources: LIME SHAP Stitch Fix Tech Blog Ravelin Blog Stripe Engineering Blog Spotify Discover Weekly Dale Carnegie’s How to Win Friends and Influence People

PyMC3 is a probabilistic programming package for Python that allows users to fit Bayesian models using a variety of numerical methods, most notably Markov chain Monte Carlo (MCMC) and variational inference (VI).

When an unbiased estimator of the likelihood is used within an Markov chain Monte Carlo (MCMC) scheme, it is necessary to tradeoff the number of samples used against the computing time. Many samples for the estimator will result in a MCMC scheme which has similar properties to the case where the likelihood is exactly known but will be expensive. Few samples for the construction of the estimator will result in faster estimation but at the expense of slower mixing of the Markov chain.We explore the relationship between the number of samples and the efficiency of the resulting MCMC estimates. Under specific assumptions about the likelihood estimator, we are able to provide guidelines on the number of samples to select for a general Metropolis-Hastings proposal.We provide theory which justifies the use of these assumptions for a large class of models. On a number of examples, we find that the assumptions on the likelihood estimator are accurate. This is joint work with Mike Pitt (University of Warwick) and Robert Kohn (UNSW).

Diffusion processes are a promising instrument to realistically model the time-continuous evolution of natural phenomena in life sciences. However, approximation of a given system is often carried out heuristically, leading to diffusions that do not correctly reflect the true dynamics of the original process. Moreover, statistical inference for diffusions proves to be challenging in practice as the likelihood function is typically intractable. This thesis contributes to stochastic modelling and statistical estimation of real problems in life sciences by means of diffusion processes. In particular, it creates a framework from existing and novel techniques for the correct approximation of pure Markov jump processes by diffusions. Concerning statistical inference, the thesis reviews existing practices and analyses and further develops a well-known Bayesian approach which introduces auxiliary observations by means of Markov chain Monte Carlo (MCMC) techniques. This procedure originally suffers from convergence problems which stem from a deterministic link between the model parameters and the quadratic variation of a continuously observed diffusion path. This thesis formulates a neat modification of the above approach for general multi-dimensional diffusions and provides the mathematical and empirical proof that the so-constructed MCMC scheme converges. The potential of the newly developed modelling and estimation methods is demonstrated in two real-data application studies: the spatial spread of human influenza in Germany and the in vivo binding behaviour of proteins in cell nuclei.

In this article we introduce a latent variable model (LVM) for mixed ordinal and continuous responses, where covariate effects on the continuous latent variables are modelled through a flexible semiparametric predictor. We extend existing LVM with simple linear covariate effects by including nonparametric components for nonlinear effects of continuous covariates and interactions with other covariates as well as spatial effects. Full Bayesian modelling is based on penalized spline and Markov random field priors and is performed by computationally efficient Markov chain Monte Carlo (MCMC) methods. We apply our approach to a large German social science survey which motivated our methodological development.

We propose a stochastic model for the analysis of time series of disease counts as collected in typical surveillance systems on notifiable infectious diseases. The model is based on a Poisson or negative binomial observation model with two components: A parameterdriven component relates the disease incidence to latent parameters describing endemic seasonal patterns, which are typical for infectious disease surveillance data. A observationdriven or epidemic component is modeled with an autoregression on the number of cases at the previous time points. The autoregressive parameter is allowed to change over time according to a Bayesian changepoint model with unknown number of changepoints. Parameter estimates are obtained through Bayesian model averaging using Markov chain Monte Carlo (MCMC) techniques. In analyses of simulated and real datasets we obtain promising results.

There has been much recent interest in Bayesian inference for generalized additive and related models. The increasing popularity of Bayesian methods for these and other model classes is mainly caused by the introduction of Markov chain Monte Carlo (MCMC) simulation techniques which allow the estimation of very complex and realistic models. This paper describes the capabilities of the public domain software BayesX for estimating complex regression models with structured additive predictor. The program extends the capabilities of existing software for semiparametric regression. Many model classes well known from the literature are special cases of the models supported by BayesX. Examples are Generalized Additive (Mixed) Models, Dynamic Models, Varying Coefficient Models, Geoadditive Models, Geographically Weighted Regression and models for space-time regression. BayesX supports the most common distributions for the response variable. For univariate responses these are Gaussian, Binomial, Poisson, Gamma and negative Binomial. For multicategorical responses, both multinomial logit and probit models for unordered categories of the response as well as cumulative threshold models for ordered categories may be estimated. Moreover, BayesX allows the estimation of complex continuous time survival and hazardrate models.

Generalized additive models (GAM) for modelling nonlinear effects of continuous covariates are now well established tools for the applied statistician. In this paper we develop Bayesian GAM's and extensions to generalized structured additive regression based on one or two dimensional P-splines as the main building block. The approach extends previous work by Lang und Brezger (2003) for Gaussian responses. Inference relies on Markov chain Monte Carlo (MCMC) simulation techniques, and is either based on iteratively weighted least squares (IWLS) proposals or on latent utility representations of (multi)categorical regression models. Our approach covers the most common univariate response distributions, e.g. the Binomial, Poisson or Gamma distribution, as well as multicategorical responses. For the first time, we present Bayesian semiparametric inference for the widely used multinomial logit models. As we will demonstrate through two applications on the forest health status of trees and a space-time analysis of health insurance data, the approach allows realistic modelling of complex problems. We consider the enormous flexibility and extendability of our approach as a main advantage of Bayesian inference based on MCMC techniques compared to more traditional approaches. Software for the methodology presented in the paper is provided within the public domain package BayesX.

Child mortality reflects a country's level of socio-economic development and quality of life. In developing countries, mortality rates are not only influenced by socio-economic, demographic and health variables but they also vary considerably across regions and districts. In this paper, we analyze child mortality in Nigeria with flexible geoadditive survival models. This class of models allows to measure small-area district-specific spatial effects simultaneously with possibly nonlinear or time-varying effects of other factors. Inference is fully Bayesian and uses recent Markov chain Monte Carlo (MCMC) simulation. The application is based on the 1999 Nigeria Demographic and Health Survey. Our method assesses effects at a high level of temporal and spatial resolution not available with traditional parametric models.

We apply a generalized Bayesian age-period-cohort (APC) model to a dataset on lung cancer mortality in West Germany, 1952-1996. Our goal is to predict future deaths rates until the year 2010, separately for males and females. Since age and period is not measured on the same grid, we propose a generalized APC-model where consecutive cohort parameters represent strongly overlapping birth cohorts. This approach results in a rather large number of parameters, where standard algorithms for statistical inference by Markov chain Monte Carlo (MCMC) methods turn out to be computationally intensive. We propose a more efficient implementation based on ideas of block sampling from the time series literature. We entertain two different formulations, penalizing either first or second differences of age, period and cohort parameters. To assess the predictive quality of both formulations, we first forecast the rates for the period 1987-1996 based on data until 1986. A comparison with the actual observed rates is made based on quantities related to the predictive deviance. Predictions of lung cancer mortality until 2010 both for males and females are finally reported.

We discuss Bayesian estimation of a logistic regression model with an unknown threshold limiting value (TLV). In these models it is assumed that there is no effect of a covariate on the response under a certain unknown TLV. The estimation of these models with a focus on the TLV in a Bayesian context by Markov chain Monte Carlo (MCMC) methods is considered. We extend the model by accounting for measurement error in the covariate. The Bayesian solution is compared with the likelihood solution proposed by Kuechenhoff and Carroll (1997) using a data set concerning the relationship between dust concentration in the working place and the occurrence of chronic bronchitis.

We present a general approach for Bayesian inference via Markov chain Monte Carlo (MCMC) simulation in generalized additive, semiparametric and mixed models. It is particularly appropriate for discrete and other fundamentally non-Gaussian responses, where Gibbs sampling techniques developed for Gaussian models cannot be applied. We use the close relation between nonparametric regression and dynamic or state space models to develop posterior sampling procedures that are based on recent Metropolis-Hasting algorithms for dynamic generalized linear models. We illustrate the approach with applications to credit scoring and unemployment duration.

This paper surveys dynamic or state space models and their relationship to non- and semiparametric models that are based on the roughness penalty approach. We focus on recent advances in dynamic modelling of non-Gaussian, in particular discrete-valued, time series and longitudinal data, make the close correspondence to semiparametric smoothing methods evident, and show how ideas from dynamic models can be adopted for Bayesian semiparametric inference in generalized additive and varying coefficient models. Basic tools for corresponding inference techniques are penalized likelihood estimation, Kalman filtering and smoothing and Markov chain Monte Carlo (MCMC) simulation. Similarities, relative merits, advantages and disadvantages of these methods are illustrated through several applications.