SUBPOPULATION INVARIANCE OF EQUATING FUNCTIONS

Abstract

This study examined population invariance of equating functions over subgroups defined by ethnicity on a teacher certification test. Investigating subgroup equating invariance was important because the total group who took this test consists of two subgroups (i.e., Hispanic and non-Hispanic) and the Hispanic group is a distinctively more able group as compared to the non-Hispanic group on the construct being measured. The study used data collected during the 2003 and 2004 administration of a teacher certification test. The chained equipercentile and linear equating methods were used to derive the equating functions. The root expected square difference was used to compare equating functions derived using the total group with equating functions derived using either the Hispanic or non-Hispanic groups. Findings suggested lack of subgroup invariance in equatings for the first test form (Form X). Also, the Hispanic group equating was less invariant as compared to the non-Hispanic group equating. The second form of the test (Form Y) showed more subgroup invariance in equating. This difference may partly be attributed to the fact that Form Y had a much bigger sample size as compared to Form X and the difference in equating functions observed in Form X for the total group and the two subgroups may be due to sampling variability. Implications of these results on actual pass/fail rates are also presented and discussed.