Journal of Educational Measurement Fall 2009, Vol. 46, No. 3, pp. 314–329 A Comparison of the Logistic Regression and Contingency Table Methods for Simultaneous Detection of Uniform and Nonuniform DIF Nes ¸e G¨ uler Sakarya University Randall D. Penfield University of Miami In this study, we investigate the logistic regression (LR), Mantel-Haenszel (MH), and Breslow-Day (BD) procedures for the simultaneous detection of both uniform and nonuniform differential item functioning (DIF). A simulation study was used to assess and compare the Type I error rate and power of a combined decision rule (CDR), which assesses DIF using a combination of the decisions made with BD and MH to those of LR. The results revealed that while the Type I error rate of CDR was consistently below the nominal alpha level, the Type I error rate of LR was high for the conditions having unequal ability distributions. In addition, the power of CDR was consistently higher than that of LR across all forms of DIF. In any assessment of individuals, test developers want to be sure that the scores are free of bias so that they are valid. Test bias occurs when the performance on the test requires sources of knowledge different from those intended to be measured, caus- ing the test scores to be less valid for a particular group (Camilli, 1993; Camilli & Shepard, 1994). The presence of item bias can be defined as the simultaneous pres- ence of two factors on item performance (Camilli, 1992; Shealy & Stout, 1993). The first factor is that the performance on a given item (called the studied item) is dependent on abilities or resources other than those intended to be measured. Us- ing the terminology of Shealy and Stout, the abilities intended to be measured are referred to as target abilities and those that are observed but unintended to be mea- sured are referred to as nuisance determinants. The second factor required for the presence of item bias is that the distribution of the nuisance determinant is unequal for two groups. By convention, the two groups are referred to as the reference and focal groups. In other words, item bias exists when the difference between focal and reference group performance on the studied item cannot be completely accounted for by differences in the target abilities, and thus must, at least partially, be accounted for by a nuisance determinant. Early investigations of item bias focused on the identification of differences in mean item scores for the focal and reference groups (Angoff, 1972; Angoff, 1982; Cleary & Hilton, 1968). It is not clear, however, whether these mean differences are attributable to a systematic bias caused by the presence of a nuisance determinant, or to a between-group difference in target ability distribution. Whether a between-group difference in mean item score is attributable to differences in the distribution of target ability or nuisance determinant can be assessed by examining between-group differ- ences in item performance after controlling for the level of target ability (Dorans & 314 Copyright c  2009 by the National Council on Measurement in Education