Impact of Multiple Sources of Correlation on Hierarchical Logistic Regression Models

• Activity Number: 169
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Katherine Cai*+ and Kyle Irimata and Jeffrey Wilson
• Companies: Arizona State University and Arizona State University and Arizona State University
• Keywords: Logistic regression ; Hierarchical model ; Mixed models ; Random effects ; Correlation ; SAS

Abstract

The analysis of hierarchical data using logistic regression presents challenges in analysis and interpretation due to the sources of correlation in each level of the data structure and any feedback between the response and covariates. The degree of correlation from each source has varying effects. Although researchers have addressed the analysis of correlated binary data, few offer guidance or understanding regarding the incorporation of random effects at each level of the hierarchy. Through simulated three-level nested data with a binary outcome, we examined methods to improve the assessment and interpretation of hierarchical data structures. We investigated the source of correlation as it pertains to the variance estimate. Although the current SAS software procedures are equipped to analyze two-level nested data, research performed present techniques to analyze higher levels.