Essays on Bayesian choice modelling with applications in health economics
The thesis develops flexible Bayesian choice models and these models are often highly-parameterised. All inference is obtained from the posterior density and evaluated using computationally intensive Markov chain Monte Carlo (MCMC) estimation. The models are applied to health economics studies and the empirical results of this thesis represent major contributions to the applied literature in their own right.The thesis consists of three separate but broadly related essays. The first essay is concerned with a heteroscedastic probit model with random effects.Real and simulated examples illustrate the approach and show that ignoring heteroscedasticity when it exists may lead to biased estimates and poor prediction. The computation is carried out by an efficient Markov chain Monte Carlo sampling scheme that generates the parameters in blocks. We use theBayes factor, cross-validation of the predictive density, the deviance information criterion, and ROC (Receiver Operating Characteristic) curves for modelcomparison. This research has been published in The Econometrics Journal. The second essay contributes to the development of Bayesian methods inthe context of valuing informal carers? needs. The random effects heteroscedastic probit model estimated in the first essay is extended to accommodate thenew data structure and account for more potential sources of heterogeneity, e.g., the individual scale effect which implies that choice behaviour is simplymore random for some people than others. The other contributions of this study include examining the entire predictive distributions of Willingness toAccept (WTA) and using the exponential transformation to allow the distribution to be skewed.The third essay follows the work of Leslie et al. (2009) on using a Dirichlet process normal mixture prior to flexibly estimate the link function of binarychoice models. This essay provides a unified framework for applying the prior to different types of contingent valuation models. We illustrate the approachthrough a multiple-bounded contingent valuation study on eliciting the Willingness to Pay (WTP) for cataract surgery in rural India. In particular wedemonstrate how to estimate an individual?s WTP distribution and its percentiles as well as the predictive probability of an affirmative answer to a givenbid price conditional on a particular set of covariates. In this essay we also conduct an experiment to examine Leslie et al.?s (2009) identification approachand find that by appropriately adjusting the magnitude of the variables, Leslie et al.?s (2009) identification approach can produce stable estimates.
Year of publication: |
2011
|
---|---|
Institutions: | Gu, Yuanyuan, Economics, Australian School of Business, UNSW ; Kohn, Robert, Economics, Australian School of Business, UNSW (contributor) ; Fiebig, Denzil, Economics, Australian School of Business, UNSW (contributor) |
Publisher: |
Awarded By:University of New South Wales. Economics |
Subject: | Dirichlete process | Discrete choice experiment | Contingent valuation |
Saved in:
freely available
Saved in favorites
Similar items by subject
-
Spatial heterogeneity of willingness to pay for forest management
Czajkowski, Mikolaj, (2015)
-
Spatial heterogeneity of willingness to pay for forest management
Czajkowski, Mikołaj, (2015)
-
Danyliv, Andriy, (2012)
- More ...