A cross-validated comparison of models for the prediction of selection, academy performance, and job tenure of police officer recruits
Date
1978
Authors
Contributor
Advisor
Department
Instructor
Depositor
Speaker
Researcher
Consultant
Interviewer
Narrator
Transcriber
Annotator
Journal Title
Journal ISSN
Volume Title
Publisher
University of Hawaii at Manoa
Volume
Number/Issue
Starting Page
Ending Page
Alternative Title
Abstract
A review of research with police populations revealed, despite the ubiquity of psychological testing and the availability of a broad range of background information, a paucity of cross-validated studies relating pre-hiring variables to selection or later performance. Although the studies reporting cross-validation have been able to account for a large portion of criterion variance, the cross-validation coefficients have been unacceptably low. To investigate the possibility that the linear regression model was not the model best suited to this sort of prediction, a variety of alternate models were explored. The following four models were selected for comparison using data from a police setting: (1) the linear model. Despite indications of the superiority of ether models with analogue and random data, and in constrained applied settings, the linear model continues to be used in virtually all predictive contexts. (2) the multiple hurdles model. This model uses subjectively-defined cutoff points on each variable. The sum of "acceptable" scores is then correlated with the criterion in a cress-validation sample. (3) the conjunctive model. This model evaluates the individuals on the minimum of their scores (modeling the situation, for instance, in which interviewers look for shortcomings on which to disqualify applicants). (4) the equal weights model. This model weights each variable equally by summing the standard scores on each to yield a predictor variate. The personnel records of 225 consecutive police officer applicants in Honolulu were searched and all demographic data were coded. A subset of the officers was administered questionnaires measuring job satisfaction, cynicism, belligerency, authoritarianism, and anomie. The demographic and attitude variables were used as predictors of selection for training, recruit school performance, and job tenure. In addition, nine of the variables were selected for cross-validated comparison of the four models. In general, it was found that the variability observed in criterion data could be well-accounted for by the linear model using the predictor variables selected. Descriptions were made of "successes" and ''failures" by each criterion. Open cross-validation, however, the coefficients showed serious shrinkage, calling into question the validity of the linear model in prediction in police personnel settings. Comparison of the alternate predictive models was also fruitful. The expected mean squared (EMS) error ratio statistic was used to compare the models. For the prediction of selection, the linear regression and multiple hurdles models appeared to be most appropriate; for the prediction of academic performance, the multiple hurdles model alone was indicated; and for the prediction of job tenure, the unit weight model appears to be as promising as any of the more complex and sophisticated models.
Description
Keywords
Police--Recruiting--Mathematical models, Police recruits, Police training
Citation
Extent
xii, 165 pages
Format
Geographic Location
Time Period
Related To
Theses for the degree of Doctor of Philosophy (University of Hawaii at Manoa). Psychology; no. 1138
Related To (URI)
Table of Contents
Rights
All UHM dissertations and theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission from the copyright owner.
Rights Holder
Local Contexts
Collections
Email libraryada-l@lists.hawaii.edu if you need this content in ADA-compliant format.