Monash University
monash_120293.pdf (1.66 MB)

An exploration of heterogeneity in ordered choice models with applications to health and health care

Download (1.66 MB)
posted on 2017-05-26, 07:53 authored by Weterings, Timothy Alan
This thesis considers a range of extensions through which heterogeneity can be included in ordered choice models, with the aim on increasing flexibility and ensuring models accurately represent the underlying data generating processes. In addition, this thesis includes discussion of variable choice in several highly relevant empirical applications including self-assessed health, tobacco consumption, and doctor utilisation. The thesis commences by offering a review of current techniques for allowing for heterogeneity. Each of the models discussed, as well as an extended model for unobserved heterogeneity, are estimated on data from the Household, Income and Labour Dynamics Survey of Australia. The second chapter investigates new extensions to unobserved threshold heterogeneity in a panel data context. The proposed model allows additional flexibility in the thresholds by permitting correlations across threshold components. The third chapter concludes analysis of self-assessed health with a full application of threshold heterogeneity. A new set of misreporting variables are considered as threshold covariates, and the model is specified to account for unobserved threshold and latent regression heterogeneity in a dynamic panel context. Chapter ef{ch:zigstop} considers another form of model flexibility, designed to account for excess observations in an outcome variable category. This is applied in a panel context to investigate issues including variable selection and unobserved heterogeneity. Chapter ef{ch:latentclass} examines latent class structure when incorporating heterogeneity more generally, allowing these classes to be ordered, and participation in each class to change over time. Each of these chapters contributes to the literature individually. However their joint contribution is to develop the quality of analysis of discrete, ordered variables in future academic research. The thesis concludes with a summary of findings, policy implications, and an outline of directions for future research.


Campus location


Principal supervisor

Mark Harris

Year of Award


Department, School or Centre

Econometrics and Business Statistics


Doctor of Philosophy

Degree Type



Faculty of Business and Economics

Usage metrics

    Faculty of Business and Economics Theses


    No categories selected


    Ref. manager