Calculation Predictor Composites in the Absence of a Criterion.
Abstract
Sometimes a prediction battery is needed for a specialty in which no adequate criterion is obtainable within a reasonable period of time. This is the second study into two techniques for producing a criterion synthetically. One technique (called R-Technique) involves the rating by experts of 100 profiles showing all the predictor variables under consideration. Another (the M-Technique) involves the rating of factors by the same experts. In both instances,the synthetic criterion data are used to produce weights for the predictors. In order to check the efficiency of the prediciton composites emerging from the use of the synthetic criteria, they are compared with a cross-validated composite obtained in the usual manner. Synthetic composites in the earlier study were predicted as well as the empirical composite; however, in this study, the results were much poorer. There are differences in predictive efficiency across the four technical schools studied, across the three sets of predictor variables used, and across the kinds of experts used (psychologists, instructors, and administrative airmen). There are also differences in efficiency associated with the size of the expert groups, with larger groups producing more effective composites. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1979
- Accession Number
- ADA080921
Entities
People
- Cecil J. Mullins
- James A. Earles
- James M. Wilbourn
Organizations
- Air Force Research Laboratory