What Will Happen to the Power If Dependent and Independent Variables Are Switched?

• Activity Number: 360
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Youngju Pak*+ and Wade Justin Davis
• Companies: University of Missouri-Columbia and University of Missouri-Columbia
• Keywords: Power ; a logistic regression ; a two sample t-test, ; binary response ; Gaussian distribution

Abstract

Researchers often face with situation where imposing a certain direction relationship between dependent (DV) and independent variables (IV) is hard to justify. Sometime a researcher can consider both directions. If researchers consider both direction i.e., switching DV and IV is possible, and if the data type of DV and IV are different, then the choice of statistical tests, thus the power also would be changed. In this paper, we are particularity interested in investigating how the power of the study changes from a logistic regression to a two sample t (or Z) test when a binary DV becomes an IV, and continuous IV becomes DV. We will provide both theoretical and empirical evidence of impact of switching DV and IV on the power of study when DV is binary and IV is assumed to follow both normal and non-normal Gaussian distribution.