ERIC Number: ED064323
Record Type: RIE
Publication Date: 1972-Apr
Reference Count: 0
The Sensitivity of Guttman Weights.
Green, Bert F., Jr.
The use of Guttman weights in scoring tests is discussed. Scores of 2,500 men on one subtest of the CEED-SAT-Verbal Test were examined using cross-validated Guttman weights. Several scores were compared, as follows: Scores obtained from cross-validated Guttman weights; Scores obtained by rounding the Guttman weights to one digit, ranging from 0 to 8; Scores obtained by approximating the Guttman weights by weights of 0, 1, and 2; Scores obtained by weighting a right answer 2, and a wrong answer 1; Scores obtained by weighting a right answer 1, and a wrong answer -- 1/4; Number right; Number wrong; and Number omitted. The reliabilities and intercorrelations of these scores are shown in tabular form. It is concluded that most of the advantages of Guttman weights come from weighting"omits" less than "wrong" answers. Because instructions in common use encourage people to omit items on a test, Guttman weighting is not recommended since with these weights the best strategy is to guess. (DB)
Publication Type: N/A
Education Level: N/A
Sponsor: National Science Foundation, Washington, DC.
Authoring Institution: Johns Hopkins Univ., Baltimore, MD. Dept. of Psychology.
Identifiers: Guttman Weighting Procedure
Note: Paper presented at the annual meeting of the American Educational Research Association (Chicago, Illinois, April 1972)