The Library
Bayesian treatment of model uncertainty under endogeneity
Tools
Miloschewski, Anne (2019) Bayesian treatment of model uncertainty under endogeneity. PhD thesis, University of Warwick.
|
PDF
WRAP_Theses_Miloschewski_2019.pdf - Submitted Version - Requires a PDF viewer. Download (1031Kb) | Preview |
Official URL: http://webcat.warwick.ac.uk/record=b3492697~S15
Abstract
The main goal of this thesis is to extend BMA methodology under endogeneity, specifically for IV models. It is comprised of five main sections. Section 2 gives an overview of the general methodology for dealing with model uncertainty. In Section 3, we review the work that has been done in this context for the standard linear regression model. From that, the reader gains a better understanding of the factors that are influential within BMA frameworks and need to be considered carefully. In Section 4, the model we base the analysis on is introduced and the effects of ignoring endogeneity within a model uncertainty framework are investigated. We now know that, not only do we obtain biased posterior point estimates (false posterior means), but also wrong posterior inclusion probabilities, if endogeneity is not accounted for properly. We then continue in Section 5 by examining a particular approach to this problem more closely. In this context, we introduce adapted prior structures in a more generalised setting for the Karl and Lenkoski (2012) approach and investigate their influence on the outcome of the estimation procedure by means of simulation studies. Additionally, we develop a tool for the classification of variables with regards to being endogenous or exogenous. To achieve that, we extended the code base of the ivbma R package. Within an empirical application example, we demonstrate that the choices of the priors do matter. Furthermore, we can point out that using our proposed prior structures and endogeneity classification approach helps to obtain improved results in terms of out-of-sample-prediction accuracy. In Section 6, we first give a short overview of Bayesian frameworks for the IV model that do not incorporate model uncertainty. Subsequently we analyse and extend the Hoogerheide et al. (2007) approach, conducting the first analytical steps towards the latter.
Item Type: | Thesis (PhD) | ||||
---|---|---|---|---|---|
Subjects: | Q Science > Q Science (General) Q Science > QA Mathematics |
||||
Library of Congress Subject Headings (LCSH): | Bayesian statistical decision theory, Uncertainty (Information theory) -- Mathematical models, Mathematical models, Mathematical statistics | ||||
Official Date: | 2019 | ||||
Dates: |
|
||||
Institution: | University of Warwick | ||||
Theses Department: | Department of Statistics | ||||
Thesis Type: | PhD | ||||
Publication Status: | Unpublished | ||||
Supervisor(s)/Advisor: | Steel, Mark F. J. | ||||
Sponsors: | University of Warwick. Department of Statistics ; Engineering and Physical Sciences Research Council | ||||
Format of File: | |||||
Extent: | 155 leaves : colour illustrations | ||||
Language: | eng |
Request changes or add full text files to a record
Repository staff actions (login required)
View Item |
Downloads
Downloads per month over past year