A Methodology to Fit Hierarchical Logistic Models for Data from Complex Sampling Designs

• Activity Number: 114
• Type: Contributed
• Date/Time: Monday, July 30, 2007 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods
• Author(s): Prabhu Bhagavatheeswaran*+ and Ian Harris
• Companies: Bristol-Myers Squibb Company and Southern Methodist University
• Keywords: Hierarchical ; Logistic ; Method of moments ; variance components

Abstract

There is an increasing interest in making inferences about the relationships among the variables investigated in a survey, as opposed to simply describing the phenomenon. Hierarchical structure of data gathering, clustering, stratification and unequal selection probabilities are taken into consideration while using linear models to make inferences on model parameters. When the response variable is binary, hierarchical logistic models are more appropriate. A modified estimation procedure based on method of moments, that is computationally less intensive , is proposed here for estimating variance components in a hierarchical logistic models .Performance of this new estimator is compared with popular estimation procedures like PQL , L-6 and GH-20. The estimator is adjusted to incorporate sampling weights that arise in modeling survey data.