Camarena Brenes, Jose Maria (2024) Generalised endogenous switching regression models and multiple imputation with applications in health economics. PhD thesis, Birkbeck, University of London.
|
Text
Brenes J final thesis for library.pdf - Full Version Download (4MB) | Preview |
Abstract
In this thesis, we present extensions of the endogenous switching regression (ESR) model with an application in health economics; and an approach to multiple imputation (MI) for variables assumed to be missing not a random. We first specify a semi-parametric ESR model where the predictors are represented using penalized regression splines, while retaining the distributional assumptions of the classical approach. We then present copula-based ESR models where the bivariate joint distributions in the model are specified using copula functions and their univariate components are specified in terms of parametric distributions. Parameter estimation and inference utilise a well-established penalized likelihood framework. We investigate insurance uptake to cover out-of-pocket expenses of prescription drugs in over 65 years olds from the United States. Our findings using the semi-parametric approach reveal evidence of self-selection into insurance and that some of the determinant factors of expenditures exhibit varying degrees of non-linear associations. An assessment of the dependence structures using the copula-based approach suggests that large values of out-of-pocket expenditures are accompanied by higher chances of having supplementary insurance however, low expenditures do not necessarily imply lower chances of having extra insurance. These features cannot be adequately captured using the classical model specification. We also present a MI approach that obtains plausible imputed values for a variable assumed to be missing not a random and not restricted to be Gaussian. The approach is derived from a copula-based specification of the sample selection model. We re-examine the non-randomised component of the REFLUX study to evaluate the effect of surgery on patient's health status under several modelling assumptions. We find that estimates of the effect of surgery are significant, regardless of the modelling approach. Estimates obtained using MI are very similar to those based on the copula model and, in some instances, they have slightly smaller standard errors.
Metadata
| Item Type: | Thesis |
|---|---|
| Copyright Holders: | The copyright of this thesis rests with the author, who asserts his/her right to be known as such according to the Copyright Designs and Patents Act 1988. No dealing with the thesis contrary to the copyright or moral rights of the author is permitted. |
| Depositing User: | Acquisitions And Metadata |
| Date Deposited: | 09 Oct 2024 16:43 |
| Last Modified: | 09 Oct 2024 19:32 |
| URI: | https://eprints.bbk.ac.uk/id/eprint/54363 |
| DOI: | https://doi.org/10.18743/PUB.00054363 |
Statistics
Additional statistics are available via IRStats2.
Archive Staff Only (login required)
