A Comparison of the Validities of a Traditional Item-Allocation Method and of An Item-Sampling Procedure in the Construction of in-Class Examinations in Intermediate Algebra

Abstract

Basically a replication of an earlier investigation by Poggio and Glasnapp (1973), this study offered from the standpoint of several psychometric characteristics a comparison of a traditional mode of item assembly with one of item sampling employed in the construction of three in-class midterm achievement examinations in intermediate algebra for a sample of 101 community college students. For two of three in-class midterm examinations reliable differences existed between mean scores of constant-item (traditionally devised) test forms and those on corresponding item-sampled (content-sampled) test forms. Relative to performance on a final achievement examination of 60 multiple-choice items, the validity coefficients of scores derived from two of the three sets of the six item-sampled test forms employed were lower than those realized for the two corresponding single constant-item test forms. Furthermore, relative to the same criterion measure multiple correlation coefficients of .571 and .521, respectively, were found from optimally weighted composites of the three sets of in-class midterm examination scores on the constant-item test forms and on the item-sampled test forms. Thus the item-sampled midterm test forms appeared to be no more and perhaps slightly less valid than traditionally assembled constant-item test forms in the prediction of success on a final objective achievement examination in intermediate algebra.