Chapter 13 - Influence of childhood experiences and present life circumstances on elderly wellbeing: A hybrid multiple ordered probit model with analytical estimation approach

Abstract

In the recent years, the integrated choice and latent variable (ICLV) modeling framework, a statistically rigorous framework that allows the appropriate treatment of the latent explanatory variables into choice models is increasingly being employed in the transportation planning arena. However, the ICLV implementations in the literature has suffered from important limitations such as simplification of the choice kernel, not allowing structural relationships between latent variables, treating indicator variables as continuous even though they are collected on a Likert scale of limited range. The identified limitations are in part related to the computational tractability issue of maximum simulated likelihood approach that serves as the workhorse for the estimation of ICLV models. In order to tap into the full potential of ICLV models it is necessary to develop alternative analytical estimation routines that do not rely on numerical simulations. The primary objective of the proposed research is to present a simulation free analytical estimation technique that allows study of multivariate choice kernels, allows hierarchical structural relationships between latent variables, and accommodates appropriate treatment of the indicator variables. The proposed estimation technique will combine composite marginal likelihood (CML) method with analytical approximation of normal cumulative density function. The ICLV model formulations and associated estimation routines will be applied on data drawn from the 2013 Disabilities and Use of Time and 2014 Childhood Retrospective Circumstances Study. The empirical study will seek to understand the influence of past experiences and present life circumstances on the elderly couples' satisfaction with life and different domains of it.