creator Kwon, Zeenatroohi advisor Morris, Scott B. Courts use different tests to evaluate adverse impact. In many cases it is beneficial to aggregate smaller samples into larger ones in order to avoid situations of low power. Aggregation through multiple events test can avoid the issue of low power and avoid statistical anomalies such as Simpson's Paradox. However, it is not always appropriate to aggregate samples because they may be too dissimilar, either through practical or statistical mean. Homogeneity of variance tests, such as the Breslow-Day test or the Modified Breslow-Day test can help the courts decide whether it is appropriate to aggregate the data or not. The present study used a Monte Carlo simulation to evaluate the performance of both tests under specific conditions similar to adverse impact analysis, while also examining under which conditions their use may be appropriate. Results showed poor power under most conditions for both tests and that the Breslow- Day test had slightly better control than the Modified Breslow-Day test under certain conditions. The tests are recommended for use when both the individual sample size and number of samples are large. Submitted by Liana Khananashvili (khananashvili@iit.edu) on 2014-10-17T18:44:12Z No. of bitstreams: 1 final thesis Z. Kwon.pdf: 616330 bytes, checksum: 3506cd44fcc1ffe2f6d206f15aedc1c6 (MD5) Made available in DSpace on 2014-10-17T18:44:12Z (GMT). No. of bitstreams: 1 final thesis Z. Kwon.pdf: 616330 bytes, checksum: 3506cd44fcc1ffe2f6d206f15aedc1c6 (MD5) Previous issue date: 2014-05 M.S. in Psychology, May 2014 2014 2014-05 http://hdl.handle.net/10560/3279 en Thesis born digital application/pdf In Copyright http://rightsstatements.org/page/InC/1.0/ Restricted Access PSYCH / Institute of Psychology Illinois Institute of Technology Affiliated department