Rossi Bayesian Non- and Semi-parametric Methods and Applications

Course Book
ISBN: 978-1-4008-5030-3
Verlag: De Gruyter
Format: EPUB
Kopierschutz: Adobe DRM (»Systemvoraussetzungen)

E-Book, Englisch, 224 Seiten

Reihe: The Econometric and Tinbergen Institutes Lectures

Peter E. Rossi is the James Collins Professor of Marketing, Economics, and Statistics at UCLA's Anderson School of Management. He has published widely in marketing, economics, statistics, and econometrics and is a coauthor of Bayesian Statistics and Marketing.

Autoren/Hrsg.

Rossi, Peter

Fachgebiete

• Wirtschaftswissenschaften Volkswirtschaftslehre Volkswirtschaftslehre Allgemein Mikroökonomie
• Wirtschaftswissenschaften Volkswirtschaftslehre Volkswirtschaftslehre Allgemein Wirtschaftstheorie, Wirtschaftsphilosophie

Inhaltsverzeichnis

Preface vii

1 Mixtures of Normals 1

1.1 Finite Mixture of Normals Likelihood Function 6

1.2 Maximum Likelihood Estimation 9

1.3 Bayesian Inference for the Mixture of Normals Model 15

1.4 Priors and the Bayesian Model 16

1.5 Unconstrained Gibbs Sampler 25

1.6 Label-Switching 29

1.7 Examples 34

1.8 Clustering Observations 46

1.9 Marginalized Samplers 49

2 Dirichlet Process Prior and Density Estimation 59

2.1 Dirichlet Processes--A Construction 60

2.2 Finite and Infinite Mixture Models 64

2.3 Stick-Breaking Representation 68

2.4 Polya Urn Representation and Associated Gibbs Sampler 70

2.5 Priors on DP Parameters and Hyper-parameters 72

2.6 Gibbs Sampler for DP Models and Density Estimation 78

2.7 Scaling the Data 80

2.8 Density Estimation Examples 81

3 Non-parametric Regression 90

3.1 Joint vs. Conditional Density Approaches 90

3.2 Implementing the Joint Approach with Mixtures of Normals 93

3.3 Examples of Non-parametric Regression Using Joint Approach 96

3.4 Discrete Dependent Variables 104

3.5 An Example of Expenditure Function Estimation 108

4 Semi-parametric Approaches 115

4.1 Semi-parametric Regression with DP Priors 115

4.2 Semi-parametric IV Models 122

5 Random Coefficient Models 152

5.1 Introduction 152

5.2 Semi-parametric Random Coefficient Logit Models 157

5.3 An Empirical Example of a Semi-parametric Random Coefficient Logit Model 161

6 Conclusions and Directions for Future Research 187

6.1 When Are Non-parametric and Semi-parametric Methods Most Useful? 187

6.2 Semi-parametric or Non-parametric Methods? 189

6.3 Extensions 191

Bibliography 195

Index 201

Über Autor(innen)

Peter E. Rossi is the James Collins Professor of Marketing, Economics, and Statistics at UCLA's Anderson School of Management. He has published widely in marketing, economics, statistics, and econometrics and is a coauthor of Bayesian Statistics and Marketing.