Modelling Unobserved Heterogeneity: Theoretical and Practical Aspects

Student thesis: Doctoral ThesisDoctor of Philosophy


The study focused on the analysis of complex survey data with hierarchical structures. Bias on estimates was studied at situations where the hierarchy was correctly or incorrectly specified using: standard logistic model as well as mixed effect models. A mathematical formula estimating bias for the maximum likelihood estimate of the logistic model was derived, and simulation was used to quantify bias in a range of situations. The study identified situations of variance of unobserved heterogeneity, where two level mixed effect models may be used instead of more complex three level models.
Date of Award1 Mar 1995
Original languageEnglish
Awarding Institution
  • King's College London

Cite this