ERIC Number: ED167590
Record Type: RIE
Publication Date: 1978-Oct
Reference Count: 0
The Statistical Structure of Multiple-Choice Items.
Donlon, Thomas F.; Fitzpatrick, Anne R.
On the basis of past research efforts to improve multiple-choice test information through differential weighting of responses to wrong answers (distractors), two statistical indices are developed. Each describes the properties of response distributions across the options of an item. Jaspen's polyserial generalization of the biserial correlation coefficient is suggested as an index of relationship between item distractors and a criterion, and a modified form of Shannon's uncertainty measure is proposed to assess the information potential of the frequency distribution across the incorrect item options. It is suggested that these indices be used to determine whether informational gains are likely to be achieved through differential weighting of item distractors. It is also suggested that they be applied to the distributions of non-right responses only, to remove the effect of the basic measurement potential of the items when dichotomously scored. Empirical results are presented which compare these indices to the difficulty and discrimination parameters of traditional item analysis, using response data from verbal and mathematical subtests of the Scholastic Aptitude Test. These item-type comparisons are used to evaluate the analytic effectiveness of the proposed statistics. Applications to further work in differential response weighting are discussed. (Author/CP)
Publication Type: Speeches/Meeting Papers; Reports - Research
Education Level: N/A
Authoring Institution: N/A
Note: Paper presented at the Annual Meeting of the Northeastern Educational Research Association (Ellenville, New York, October, 1978)