NCJ Number
63271
Date Published
1979
Length
42 pages
Annotation
THE VULNERABILITY OF RANDOM TIME QUOTA SELECTION TO SELECTION BIAS IN THE IMPLEMENTATION OF AN EVALUATION DESIGN IS EXAMINED USING MARKOV METHODS OF ANALYSIS.
Abstract
ALTHOUGH THE RANDOM ASSIGNMENT OF CLIENTS TO EXPERIMENTAL AND CONTROL GROUPS IS GENERALLY ACCEPTED AS THE EVALUATION DESIGN MOST LIKELY TO YIELD UNEQUIVOCAL RESULTS, ETHICAL AND LEGAL CONSIDERATIONS, NOTABLY THE PROPRIETY OF RANDOMLY DENYING A PRESUMABLY BENEFICIAL TREATMENT TO THE CONTROL GROUP OR POSSIBLY SUBJECTING THE EXPERIMENTAL GROUP TO A POTENTIALLY HARMFUL INFLUENCE, PRESENT OBSTACLES TO SUCH A DESIGN. THE VERA INSTITUTE, IN ITS EVALUATION OF THE NEW YORK COURT EMPLOYMENT PROJECT, USED 'RANDOM TIME QUOTA SELECTION' AS AN ACCEPTABLE ALTERNATIVE TO RANDOM ASSIGNMENT. UNDER THIS METHOD, THE TOTAL DURATION OF THE EVALUATION WAS DIVIDED INTO SHORT TIME PERIODS (1 TO 3 WORKING DAYS) OF RANDOM LENGTH. THEN, PROPORTIONAL TO THE LENGTH OF EACH TIME PERIOD, A QUOTA WAS ESTABLISHED FOR THE MAXIMUM NUMBER OF DEFENDANTS WHO COULD BE ACCEPTED INTO THE PROGRAM (EXPERIMENTAL GROUP) DURING THAT PERIOD. DEFENDANTS DEEMED ELIGIBLE TO RECEIVE PROGRAM SERVICES BUT WHO WERE NOT ACCEPTED UNDER THE CRITERION FORMED THE CONTROL GROUP. USING THE TECHNIQUES OF MARKOV ANALYSIS TO INVESTIGATE THE BEHAVIOR OF A SIMPLIFIED SYSTEM OF RANDOM TIME QUOTA SELECTION, THE RESEARCHERS PRODUCED RESULTS SHOWING THAT, WITH PROPER CARE, THE METHOD CAN BE USED AS AN ALTERNATIVE TO RANDOM SELECTION WITHOUT INCURING UNDUE SELECTION BIAS. MATHEMATICAL EQUATIONS AND REFERENCES ARE PROVIDED. (RCB)