Comparing Regression Equations Across Training Programs: An Empirical Study of Prior Selection Effects and Alternative Prediction Composites.
Abstract
This study examines the possible magnitudes of incidental selection effects on prediction equations for performance in twenty-seven technical training programs in the Marine Corps. Selection composites for Forms 8, 9, and 10 of the Armed Forces Vocational Aptitude Battery (ASVAB) are shown to yield widely varying predictions of final grades in training when the usual least-squares regression equations are examined. These predictions are considerably more homogeneous and consistent with course content when a procedure for simultaneous estimation of regression coefficients is used and when regression coefficients are adjusted for incidental selection. Implications of these findings for the development of alternative prediction composites are discussed and illustrated by examining alternatives to the present ASVAB composite for selecting trainees in clerical programs. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 07, 1986
- Accession Number
- ADA169616
Entities
People
- Stephen B. Dunbar
Organizations
- University of Iowa