Sources of Correlation on a Hierarchical Logistic Regression Model

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

Abstract

The correlation present in a hierarchical structure is due to different types of levels of associations. There is the correlation on account of repeated measures on the sampling unit, the feedback effect from the response to the covariate, and the correlation among the covariates. There are a variety of methods available which take this correlation into account; however, these methods usually address the correlations separately and rarely address them simultaneously. Further, there is a lack of information regarding the amount of correlation that will cause issues in the analysis of data from any of these sources. In this study we seek to determine how each source of correlation derived affects the significance of the predictors on the outcome collectively and separately. We produce a measure of correlation which provides information regarding the overall level of association. We also establish a threshold at which the correlation due to the repeatedness will begin to affect the significance of the covariate. This threshold is useful to identify situations in which researchers should explore the use of models which take the correlation strength within their data into account.