Conditional Logistic Regression with Survey Data

• Activity Number: 254
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Barry I. Graubard*+ and Edward L. Korn
• Companies: National Cancer Institute and National Cancer Institute
• Keywords: Conditional logistic regression ; complex sample surveys ; pseudo pairwise likelihood ; Hispanic Health and Nutrition Examination Survey

Abstract

Conditional logistic regression (CLR) of cluster samples is used to relate a binary outcome to covariates conditionally on the cluster effects. Surveys can use multistage sampling with differential weighting of individuals from the conditioning cluster (e.g. family). We show that CLR of survey data using inflation (weighted) estimation can be biased for small sample sizes within conditioning clusters. We propose methods using weighted pseudo pairwise likelihoods that combine the CLR likelihoods for all pairs of observations with a positive and negative outcome within a cluster and weights the pairwise likelihoods by the inverse of their joint inclusion probabilities within the cluster. Variance estimators are provided. Simulations show that the proposed methods are approximately unbiased. An example from the Hispanic Health and Nutrition Examination Survey is used for illustration.