EQUATING ERROR AND STATISTICAL BIAS IN SMALL SAMPLE LINEAR EQUATING

A resampling study was conducted to compare the statistical bias and s tandard errors of nonequivalent-groups linear test equating in small s amples of examinees. Sample sizes of 15, 25, 50, and 100 were examined . One thousand samples of each size were drawn with replacement from e ach of 5 archival data files from teacher subject area tests. For each test, data files from 2 parallel forms were used. Results suggest tri vial levels of equating bias even with small samples, but substantial increases in standard errors as sample size decreases. Results were in terpreted in terms of applications to testing situations in which smal l numbers of examinees are available.