A differential item functioning (DIF) simulation study was conducted to explore the type and level of impact that contamination had on type I error and power rates in DIF analyses when the suspect item favored the same or opposite group as the DIF items in the matching subtest. Type I error and power rates were displayed separately for the reference and focal groups, using both statistical test and statistical test with DIF magnitude criteria, to compare the differential impacts on each group of varying contamination conditions in the matching subtest. Simulated study conditions were created by manipulating the percentage of items in the matching subtest that contained DIF (0%. 10%, 20%, and 30%), magnitude of DIF in the matching subtest DIF items and for the suspect item (d[subscript eta] = 0.5, 1.0, and 1.5), and sample size and sample size ratio (N[subscript R]:N[subscript F] = 500:500, 1000:1000, 750:250, and 1500:500). DIF analyses were conducted using SIBTEST. The major statistical outcome was that while power decreased when investigating a suspect item that favored the same group as the DIF items in the matching subtest, power actually increased when a suspect item favored the group not favored by DIF items in the matching subtest. The results indicated that DIF analyses conducted on tests with unbalanced numbers of total DIF items can result in an under-identification of DIF for the group most favored in the test and an over-identification of DIF items for the group least favored by test items. Implications from this research are that some purification procedures could be creating more biased and not less biased matching subtests, potentially impacting DIF item interpretation and item selection. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtmlTest.]
