NCJ Number
52521
Journal
Journal of Consulting and Clinical Psychology Volume: 45 Issue: 4 Dated: (AUGUST 1977) Pages: 559-563
Date Published
1977
Length
5 pages
Annotation
THREE LINEAR SCALES WERE COMPARED WITH FOUR CONFIGURAL STRATEGIES IN THE PREDICTION OF RECIDIVISM AMONG ADULT MALE PROBATIONERS IN DALLAS COUNTY, TEX.
Abstract
A SAMPLE OF 153 NONREVOKED PROBATIONERS WAS SELECTED FROM A TOTAL OF 2,041 MALES PLACED ON PROBATION BETWEEN JANUARY 1, 1973, AND OCTOBER 31, 1973, FOR CONVICTION OF A STATE FELONY. CASE FILES OF 150 REVOKED PROBATIONERS AND THE 153 NONREVOKED PROBATIONERS WERE RANDOMLY ASSIGNED TO A DERIVATION SAMPLE AND A VALIDATION SAMPLE. THE DERIVATION SAMPLE CONSISTED OF 75 REVOKED CASES AND 77 NONREVOKED CASES. THE VALIDATION SAMPLE WAS COMPOSED OF 75 REVOKED CASES AND 76 NONREVOKED CASES. DESCRIPTIVE INFORMATION WAS RECORDED FOR ALL PROBATIONERS, INCLUDING 37 DICHOTOMOUS DESCRIPTORS, 1 POLYTOMOUS DESCRIPTOR (TYPE OF CURRENT OFFENSE), AND 1 CONTINUOUS VARIABLE (AGE WHEN PLACED ON PROBATION). ONE LINEAR AND FOUR CONFIGURAL PREDICTION STRATEGIES WERE DEVELOPED USING THE DERIVATION SAMPLE AND TESTED WITH THE VALIDATION SAMPLE: CUMULATIVE POINTS ANALYSIS, PREDICTIVE ATTRIBUTE ANALYSIS, MEAN COST RATING ANALYSIS, PATTERN PROBABILITY ANALYSIS, AND CONFIGURATION ANALYSIS. LINEAR SCALES WERE ALL SUPERIOR TO THE CONFIGURAL STRATEGY WHEN COMPARED USING CHI-SQUARE TECHNIQUES, TOTAL HIT RATE, AN INDEX OF THE SPREAD OF FAILURE RATES AMONG RISK GROUPS, AN INDEX OF REDUCTION IN CLASSIFICATORY CRIME, AND AN INDEX OF THE STABILITY OF FAILURE RATES BETWEEN DERIVATION AND VALIDATION GROUPS. REQUIREMENTS FOR THE COMPREHENSIVE EVALUATION OF PREDICTION STRATEGIES ARE DISCUSSED. SUPPORTING DATA AND REFERENCES ARE INCLUDED. (DEP)