Consumer expenditure estimation incorporating generalized variance functions in hierarchical Bayes models

Abstract

This thesis investigates models for sampling error that would improve the stability of mean annual consumer expenditure estimates. Hierarchical Bayes models presented consider mean estimates from five interview and diary sources for diary only estimates when interview estimates are also available. The investigation breaks down the problem into improving the generalized variance function estimate of variance through three-digit, four-digit and five-digit groupings of universal classification codes, estimation of sampling variance-covariance using hierarchical models and mean expenditure estimation.