Inating It Points of View or Opin

Because previous reports have suggested that the lowered validity of tests scored with empirical option weights might be explained by a capitalization of the keying procedures on omitting tendencies, a procedure was devised to key options empirically with a "correction-for-guessing" constraint. Use of the new procedure with Graduate Record Examinations (GRE) data resulted in smaller increases in reliability than those observed when unconstrained procedures were used, but validities for quantitative subforms were not appreciably lowered. Validities for verbal subforms were lowered slightly, however. Empirical Option Weighting with a Correction for Guessing 1 Richard R. Reilly Educational Testing Service Two recent reports (Hendrickson, 1971; Reilly & Jackson, 1972) have suggested that weighting options empirically results in substantial increases in reliability and test homogeneity, but at the expense of lowered test validity. ThEse findings are at variance with those reported in an earlier study by Davis and Fifer (1959) who found similar increases in reliability and slight increases in validity when options were weighted empirically. All three studies employed modifications of a weighting technique originally known as The Method of Reciprocal Averages (Mosier, 1946) which, in effect, maximizes the product-moment correlatior etween item scores and criterion scores by assigning to each item-option values proportional to the mean criterion score for all individuals chaling that option. A key difference between the Davis and Fifer study and the first two mentioned was that testa in the first two were administered with formula score instructions while Davis and Fifer instructed examinees to attempt every item. Thus, Hendrickson and Reilly and Jackson had an additional "option," that of omit. Fi ndrickson, reporting on the weights generally a-signed to the omit category comments, "...An interesting finding of this study was that the weight of 'omit' was almost always lower than any of the other distracters in an item..." (Hendrickson, 1971). Reilly and Jackson (1972) take this J, step further and suggest that, "...the empirical keying procedures described capitalize on the tendency to omit and...while this tendency is reliable, it is not valid."