CONTRAST: A COMPUTER PROGRAM FOR EVALUATING ITEM BIAS

Described is a program for a comparative evaluation of the extent of item bias between two subgroups in a test population. The pro-gram calculates an index of bias based on Angoff’s elliptical distance measure as well as provides statistics for determining the similarity of intergroup item parameters. Also available are options for rejecting items on the basis of parameter tolerances and for labelling program output with a variable length of text. EDUCATIONAL AND PSYCHOLOGICAL MEASUREMENT 1978, 38 THERE is currently a heightened awareness among test developers and researchers of the problem of bias in mental ability tests. Briefly stated, this problem is that items in some of the tests used to assess mental ability may be measuring attributes which reflect ethnic or cultural background instead of, or in addition to, measuring mental ability. Test items of this kind are &dquo;biased&dquo; in the sense that including them in a test of mental ability will tend to favor some subgroups of the general population rather than others. Numerous techniques have been developed to detect biased items. (See, for example, the reviews by Angoff, 1975). Most of these tech-niques involve employing either factor analysis to determine whether a given test measures the same attributes for two or more subgroups, or some method for detecting a subgroup by item parameter interaction