271 – SPEED: Recent Advance of Statistical Methods in Biometrics, Part 2
Estimating Power for Interaction Tests in Logistic Regression: A Case Study of Tobacco Cessation Among Cancer Survivors
Zoran Bursac
University of Tennessee Health Science Center
D. Keith Williams
University of Arkansas for Medical Sciences
C. Heath Gauss
University of Arkansas for Medical Sciences
Robert Klesges
University of Tennessee Health Science Center
At the present time, researchers are limited in available methods to conduct power analysis for an interaction term between two main variables of interest in a study that utilizes logistic regression. We propose a method and a SAS macro tool for estimating the power associated with an interaction term in a logistic regression model. This method empirically calculates the power for an interaction term, based on cell counts from a 2x2x2 table, and several other intuitive input parameters. We illustrate the method with an example from a randomized controlled trial (RCT) of tobacco cessation among cancer survivors, which investigates interaction between two-level treatment assignment and cancer staging.