A new statistical method will be evaluated for detecting differential functioning of test items for subgroups of people. Application will ensure accurate unbiased assessment of cognitive functioning, which is important in assessing a person's mental and physical state. If a cognitive test is biased for or against some group of persons, the assessment will be flawed. The new method uses log-linear analysis of three-way contingency tables to determine if particular wrong choices on an item are differentially attractive to different groups, when ability is controlled. Significantly group differences could indicate that the item had different meaning to members of different groups. Analyses of available data from the Scholastic Aptitude Test and analyses of simulated data will permit needed evaluation of the new method.
