Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.386202
Title: Event history analysis : discrete-time models including unobserved heterogeneity, with applications to birth history data
Author: Egger, Peter Johann
ISNI:       0000 0001 3441 8809
Awarding Body: University of Southampton
Current Institution: University of Southampton
Date of Award: 1992
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Abstract:
Event histories are records on the durations between events which indicate transitions from one state to another. Event history models focus on the analysis of these durations. There are several situations where discrete time is the appropriate time scale. Only a modest amount of work has been done on event history models in discrete time. Different approaches for modelling event histories in discrete time are explored in this work. Some of these methods are extended to analyse observed heterogeneity by modelling covariates. Special attention is given to unobserved heterogeneity in the risk of having an event. If unobserved heterogeneity is not taken into account, the results in the event history models are biased. In this work unobserved heterogeneity is modelled through extra-random effects. The simplest event history model in discrete time is the geometric model. Since geometric event history data can also be considered as a series of Bernoulli trials, these event history data can also be analysed through logistic regression models. Thus, unobserved heterogeneity can be modelled through random-effects logistic regression models using standard software. We compare the geometric model and two of these models, namely the beta-geometric model and the logistic-normal (-geometric) model, using simulations. The logistic-normal model can also be interpreted as a multilevel model for the analysis of hierarchically structured data. The logistic-normal model is used to analyse Scottish birth history data. Special attention is given to the proximate determinant use of contraception, where the effects of ignoring the length of use of contraception are analysed and its implications discussed. Finally, the random-effects logistic models are applied to the analysis of the Hutterite birth history data. The Hutterites are a natural fertility population, well known for its high fertility.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.386202  DOI: Not available
Keywords: Statistics
Share: