Bayesian Smoothing and Regression for Longitudinal, Spatial and Event History Data
, by Fahrmeir, Ludwig; Kneib, ThomasNote: Supplemental materials are not guaranteed with Rental or Used book purchases.
- ISBN: 9780199533022 | 0199533024
- Cover: Hardcover
- Copyright: 6/24/2011
Several recent advances in smoothing and semiparametric regression are presented in this book from a unifying, Bayesian perspective. Simulation-based full Bayesian Markov chain Monte Carlo (MCMC) inference, as well as empirical Bayes procedures closely related to penalized likelihood estimation and mixed models, are considered here. Throughout, the focus is on semiparametric regression and smoothing based on basis expansions of unknown functions and effects in combination with smoothness priors for the basis coefficients. Beginning with a review of basic methods for smoothing and mixed models, longitudinal data, spatial data and event history data are treated in separate chapters. Worked examples from various fields such as forestry, development economics, medicine and marketing are used to illustrate the statistical methods covered in this book. Most of these examples have been analyzed using implementations in the Bayesian software, BayesX, and some with R Codes.
Ludwig Fahrmeir is Professor Emeritus, Department of Statistics, Ludwig-Maximilians-University Munich. He has been Professor of Statistics at the University of Regensburg, Chairman of the Collaborative Research Centre "Statistical Analysis of Discrete Structures with Applications in Econometrics and Biometrics" and was coordinator of the project "Analysis and Modelling of Complex Systems in Biology and Medicine" at the University of Munich. He is an Elected Fellow of the International Statistical Institute.
Thomas Kneib received a PhD in Statistics in 2006 from the University of Munich. He has been visiting Professor for Applied Statistics at the University of Ulm and Professor for Statistics at the University of Gottingen. Currently, he is Professor for Applied Statistics at the University of Oldenburg.
List of Algorithms | p. XV |
List of Examples | p. xvi |
Introduction: Scope of the Book and Applications | p. 1 |
Semiparametric regression | p. 1 |
Applications | p. 4 |
Basic Concepts for Smoothing and Semiparametric Regression | p. 18 |
Time series smoothing | p. 19 |
Gaussian observation models | p. 19 |
Penalized least-squares smoothing | p. 20 |
Bayesian smoothing | p. 23 |
Some modifications and extensions | p. 27 |
Estimation of smoothing parameters and variances | p. 27 |
Other model components | p. 27 |
Correlated errors | p. 28 |
Locally adaptive smoothing | p. 29 |
Unequally spaced time-series observations | p. 30 |
Non-Gaussian observation models | p. 30 |
Semiparametric regression based on penalized splines | p. 34 |
Gaussian observation models | p. 34 |
Polynomial splines | p. 35 |
Truncated power series and B-splines | p. 36 |
Nonparametric regression based on polynomial splines | p. 39 |
Characteristics of a spline fit | p. 41 |
P-splines | p. 42 |
Customized penalties | p. 47 |
Bayesian P-splines | p. 48 |
Bayesian inference | p. 52 |
Degrees of freedom of a P-spline | p. 55 |
Univariate non-Gaussian observation models | p. 57 |
Penalized likelihood estimation | p. 58 |
Bayesian inference | p. 60 |
Latent variable representations | p. 63 |
Categorical observation models | p. 67 |
Nominal response models | p. 68 |
Cumulative models for ordinal responses | p. 73 |
Sequential models for ordinal responses | p. 77 |
Penalised likelihood inference | p. 79 |
Bayesian inference | p. 81 |
Related smoothing approaches | p. 82 |
Integral penalties | p. 82 |
Smoothing splines | p. 83 |
Bayesian interpretation of smoothing splines | p. 85 |
Reproducing kernel Hilbert spaces | p. 87 |
Other types of basis functions | p. 89 |
Generalized additive models | p. 90 |
Gaussian additive models | p. 90 |
Simultaneous penalized least-squares (PLS) smoothing | p. 92 |
Backfitting | p. 93 |
Bayesian backfitting: the Gibbs sampler | p. 94 |
Non-Gaussian additive models | p. 100 |
Notes and further reading | p. 104 |
Generalized Linear Mixed Models | p. 107 |
Linear mixed models with Gaussian random effects | p. 108 |
Linear mixed models for longitudinal data | p. 108 |
Advantages of mixed models | p. 111 |
Marginal and conditional formulation | p. 114 |
Multilevel models | p. 115 |
General linear mixed models | p. 116 |
Bayesian linear mixed models | p. 117 |
Likelihood-based inference | p. 119 |
Estimation and prediction | p. 119 |
Estimation of regression coefficients for given variance components | p. 121 |
Maximum likelihood (ML) estimation of variance components | p. 122 |
REML estimation for the variance parameters | p. 122 |
Details on REML estimation for variance parameters | p. 124 |
Bayesian interpretation of ML and REML estimation | p. 125 |
Testing hypotheses | p. 126 |
Bayesian inference | p. 127 |
Empirical Bayes inference | p. 127 |
Full Bayes inference | p. 128 |
Full Bayes inference for longitudinal data | p. 129 |
Linear mixed models with flexible random effects priors | p. 138 |
Finite mixture models | p. 139 |
Finite mixture of normals prior: the heterogeneity model | p. 140 |
MCMC inference for finite mixture models | p. 141 |
Penalized mixture of normals priors | p. 142 |
Dirichlet processes | p. 143 |
Dirichlet process: descriptive definition | p. 145 |
Stick breaking representation | p. 145 |
Posterior Dirichlet process | p. 147 |
Predictive distribution and clustering property | p. 148 |
Dirichlet process mixtures | p. 149 |
LMM with DP-based random effects priors | p. 150 |
Longitudinal data LMMs with DP random effects priors | p. 150 |
Longitudinal data LMMs with DPM priors | p. 153 |
Generalized linear mixed models | p. 155 |
Generalized linear mixed models for longitudinal data | p. 157 |
GLMMs for univariate responses | p. 157 |
Marginal and conditional models | p. 159 |
Interpretation of regression parameters | p. 160 |
GLMMs for categorical responses | p. 161 |
General mixed models for non-Gaussian responses | p. 164 |
Likelihood-based and empirical Bayes inference | p. 164 |
Full Bayesian inference for longitudinal data | p. 171 |
Longitudinal data GLMMs with flexible random effects priors | p. 174 |
GLMMs with DP random effects priors | p. 174 |
GLMMs with DPM random effects priors | p. 175 |
Notes and further reading | p. 176 |
Semiparametric Mixed Models for Longitudinal Data | p. 178 |
Semiparametric mixed models based on Gaussian priors | p. 179 |
Observation models for univariate responses from exponential families | p. 179 |
Generalized additive mixed models | p. 179 |
Varying coefficient mixed models | p. 181 |
ANOVA type interactions | p. 182 |
Generic representation | p. 182 |
Observation models for categorical responses | p. 183 |
Gaussian priors for regression parameters and functions | p. 185 |
Estimation | p. 187 |
Empirical Bayes inference | p. 187 |
Intuitive example | p. 187 |
Mixed model representation for P-splines | p. 190 |
Mixed model representation for general penalized smoothers | p. 192 |
Mixed model-based estimation of SPMMs | p. 196 |
Identifiability | p. 200 |
Constrained smoothness priors | p. 205 |
Credible intervals and bands | p. 206 |
Tests on the functional form | p. 207 |
Full Bayes estimation with Gaussian priors for regression parameters | p. 212 |
Gaussian SPMMs | p. 213 |
Exponential family SPMMs | p. 219 |
Categorical SPMMs | p. 222 |
Credible intervals and bands | p. 225 |
Smoothing and correlation | p. 229 |
Correlations induced by penalized smoothing approaches | p. 230 |
Identifiability problems | p. 233 |
Radial basis functions and correlated errors | p. 237 |
Summary | p. 238 |
Extensions based on non-Gaussian priors | p. 238 |
SPMMs with DP-based random effects priors | p. 238 |
Shrinkage priors for high-dimensional regression parameters | p. 245 |
Ridge prior | p. 245 |
Lasso prior | p. 247 |
Lq priors | p. 254 |
Further examples | p. 255 |
Locally adaptive priors for functions | p. 256 |
Locally adaptive penalties | p. 256 |
Knot selection strategies | p. 261 |
Model choice and model checking | p. 263 |
Bayes factors and model selection criteria | p. 264 |
Bayes factors and marginal likelihoods | p. 264 |
Methods for estimating Bayes factors and marginal likelihoods | p. 266 |
Information criteria: AIC, BIC and DIC | p. 269 |
BIC | p. 270 |
AIC | p. 270 |
DIC | p. 275 |
Predictive methods for model assessment | p. 276 |
Alternative predictive distributions | p. 277 |
Assessing calibration: Pit and Bot | p. 278 |
Proper scoring rules | p. 281 |
Custom summary statistics | p. 282 |
Monte Carlo estimation of predictive measures | p. 283 |
Posterior predictive goodness-of-fit assessment | p. 284 |
Exact cross validatory predictive assessment | p. 286 |
Approximate cross validatory predictive assessment | p. 287 |
Predictor selection using spike and slab priors | p. 290 |
Variable selection | p. 291 |
Function selection | p. 297 |
Covariance matrix selection for random effects | p. 302 |
Notes and further reading | p. 303 |
Individual-specific curves and functional mixed models | p. 303 |
Approximate Bayesian inference | p. 304 |
Variational Bayes approaches | p. 304 |
Integrated nested laplace approximation (INLA) | p. 305 |
Further comments | p. 306 |
Spatial Smoothing, Interactions and Geoadditive Regression | p. 307 |
Spatial data structures | p. 309 |
Point-referenced data: Continuous spatial information | p. 309 |
Interaction surfaces | p. 314 |
Areal data: Discrete spatial information | p. 318 |
Continuous vs. discrete spatial information | p. 321 |
Spatial regression models | p. 322 |
Other types of spatial data | p. 325 |
Discrete spatial data: Markov random fields | p. 325 |
A heuristic spatial smoothness prior | p. 326 |
Markov random fields | p. 328 |
Definition of Markov random fields | p. 328 |
Brook's lemma | p. 330 |
Negpotential function and Hammersley-Clifford theorem | p. 332 |
Auto-models | p. 334 |
Gaussian Markov random fields/auto-normal models | p. 336 |
Basis function representation of Gmrfs | p. 340 |
Direct vs. latent autoregressive models | p. 344 |
Extended Markov random field models | p. 346 |
Spatial smoothing approaches and interactions | p. 347 |
Tensor product penalized splines | p. 347 |
Tensor product bases | p. 348 |
Kronecker product penalties for tensor product bases | p. 352 |
Generalized penalty concepts | p. 358 |
Null spaces of bivariate penalties | p. 362 |
Higher-order Markov random fields on regular grids | p. 364 |
Higher-order interactions | p. 364 |
Radial bases | p. 366 |
Space filling algorithm | p. 370 |
Penalties for radial bases | p. 371 |
Tensor products vs. radial bases | p. 373 |
Continuous spatial data: Stationary Gaussian random fields | p. 374 |
Model formulation | p. 375 |
Correlation functions | p. 378 |
Parametric classes of correlations functions | p. 381 |
Bochner's theorem | p. 385 |
Range anisotropic correlation functions | p. 386 |
Variogram | p. 389 |
Estimation of covariance and correlation parameters | p. 391 |
Nonparametric covariogram and variogram estimation | p. 392 |
Gaussian random fields as radial basis function smoothers | p. 395 |
Identifiability in geostatistical models | p. 396 |
Classical geostatistics | p. 398 |
Geoadditive regression | p. 399 |
Full Bayes inference | p. 400 |
Empirical Bayes inference | p. 407 |
Notes and further reading | p. 413 |
Event History Data | p. 415 |
Survival data | p. 419 |
Basic notions for continuous survival times | p. 419 |
Censoring and truncation | p. 420 |
Likelihood contributions for different types of censoring | p. 424 |
Discrete-time survival data | p. 425 |
Continuous-time hazard regression | p. 428 |
Observation models, priors and likelihoods | p. 430 |
Observation models | p. 430 |
Priors | p. 431 |
Likelihoods | p. 432 |
Full Bayes inference for right-censored observations | p. 433 |
Piecewise exponential model | p. 433 |
Models with general structured additive predictor | p. 434 |
Empirical Bayes (EB) inference for right-censored observations | p. 444 |
Inference for interval-censored observations | p. 451 |
Discrete-time hazard regression | p. 455 |
Accelerated failure time models | p. 464 |
Observation models, likelihoods and priors | p. 466 |
Penalized Gaussian mixture (PGM) | p. 467 |
Aft models with DP(M) priors | p. 470 |
Multi-state models | p. 470 |
Continuous-time transition rate models | p. 471 |
Counting process representation and likelihood Contributions | p. 472 |
Empirical and full Bayes inference | p. 474 |
Model checking based on martingale residuals | p. 481 |
Discrete-time multi-state models | p. 485 |
Notes and further reading | p. 488 |
Models for correlated survival data | p. 490 |
Joint modelling of longitudinal and event history data | p. 492 |
Bibliography | p. 495 |
Index | p. 519 |
Table of Contents provided by Ingram. All Rights Reserved. |
What is included with this book?
The New copy of this book will include any supplemental materials advertised. Please check the title of the book to determine if it should include any access cards, study guides, lab manuals, CDs, etc.
The Used, Rental and eBook copies of this book are not guaranteed to include any supplemental materials. Typically, only the book itself is included. This is true even if the title states it includes any access cards, study guides, lab manuals, CDs, etc.