AN EMPIRICAL COMPARISON OF FIVE METHODS OF SHORTENING A TEST*
Marjorie Olsen, William B. Schrader · ETS Research Bulletin Series · 1953
ABSTRACT This study was designed to provide an empirical comparison of the following methods of shortening the Law School Admission Test: (1) Internal consistency item analysis; (2) External criterion item analysis; (3) A method involving the ratio of item‐criterion correlation to item‐subtest correlation, substantially as described by Gulliksen; (4) A method described by Horst for estimating optimal times for subtests; and (5) A linear method for estimating optimal times for subtests. The analysis was carried out separately for two parallel subsamples of 500 first‐year law school students, and trial scoring keys requiring 135 minutes of testing time were assembled from material originally requiring 190 minutes. When the trial scoring keys obtained from one subsample were validated against first‐year law school grades for the opposite subsample, the results were as follows: For one subsample, the validity coefficients for the five trial keys ranged from .568 to .576; the coefficient for total score on the 190 minute test was .586. For the second subsample, the validities of the trial keys were between .483 and .496, as compared with .493 for the 190 minute test.