The Comparative Validities of Traditional Item-Selection and Item-Sampling Procedures in Evaluation of Student and Class Progress Toward Course Objectives in Freshman Psychology

Abstract

Constituting essentially a replication of an earlier investigation by Poggio and Glasnapp (1973), this study was directed toward assessing the comparability of test scores arising from use of traditional procedures of assembling test items to prepare constant-item test forms with those scores associated with employment of content-sampling (matrix-sampling) techniques to construct item-sampled test forms. The sample was 104 community college students in northern California, who were enrolled in a one-semester introductory psychology course. Comparisons were made in the mean scores on the two types of test forms employed in each of three in-class midterm examinations as well as of the validity coefficients of these two types of test forms relative to each of two criterion measures representing two parts of a final achievement examination. Although evidence regarding the comparability of the score distributions of the two kinds of test forms was equivocal, item-sampled test forms were no more valid than were constant-item test forms. Moreover, optimally weighted composites of three item-sampled test forms were no more valid than were corresponding composites of three constant-item test forms.