This is a report on some considerations involved in validating an occupational test battery for a racially mixed group. The 75% white and 25% black sample showed clear-cut differences in performance on the predictor variables and ambiguous differences in performance on the criterion variables. Cross-validations demonstrated that poorest prediction coefficients (sometimes 0 or negative) resulted when predicting across the racial groups using the same prediction equation for members of both groups. Predictions improved for both groups when race-specific equations were used, the best predictions being obtained for the black group. Analysis of the beta weights of the most effective predictor variables pinpointed the attributes common to the more successful, as distinguished from the less successful, of the currently employed black and white patrolmen.

No abstract

The use of supervisory assessments of performance during the course of a test validation study raised statistical problems in the handling of variable degrees of racial bias on the part of the raters and differing lengths of job tenure on the part of the employees. This paper describes a method of solving these problems which permitted use of 100% of the available supervisory ratings.

In the course of validating a test battery for the selection of police officers (Baehr, Furcon, & Froemel, 1969), project personnel were searching for objectively scored, pencil-and-paper measures of behavior variables. One product of this search was an adaptation of the Arrow-Dot subtest of the four-part IES test battery (Dombrose & Slobin, 1958). This adaptation was designed for group administration and for improved reliability as a possible predictor of performance. Intercorrelational analysis and factor analysis of intercorrelations with other potential predictor variables and performance criteria provided some information on the nature of the Arrow-Dot scores. The test's ability to predict police officers' performance was assessed through multiple regression analyses of its scores both as independent predictors and as contributors of unique variance in a wider predictor battery. Its usefulness for diagnosis of specific problems in police officers' performance was also investigated. Application in five additional studies permitted an exploration of the “characteristic” Arrow-Dot dimension profile for police officers.

No abstract