Document Type

Thesis - Open Access

Award Date


Degree Name

Doctor of Philosophy (PhD)


Sociology and Rural Studies


The study investigated the following problem: What is the best estimator of failure to appear (FTA) for felons in Ramset County, Minnesota, who are released into the community during their pretrial period? An estimator of FTA and nonFTA probabilities was derived with the aid of stepwise discriminant analysis. The method demonstrates the expost facto ability to correctly place 98 percent of the nonFTA’s and 30 percent of the FTA’s in the 1975 sample. The overall total group placement, due to the large percentage of nonFTA’s in the sample, was approximately 90 percent. Even though it is recognized that multiple regression is not the most appropriate technique to use with a dichotomized dependent variable, a comparison was made between the two statistical methods. The total R2 in the complete estimator accounted for a little over 14 percent of the total variance. Multiple regression was used as a comparative method along with discriminant analysis because of its popularity and not because of its applicability. Discriminant analysis is the preferred method in this study because: (1) it is recommended when the researcher has a large number of variables that appear to be highly correlated, (2) it constructs a weighted sum, or linear combination of the major original discriminating variables, and (3) it produces a probability of group membership for both the FTA and nonFTA samples. Findings from this research suggest that (1) the Vera Scale is inappropriately weighted for use as a release instrument in Ramsey County, Minnesota, (2) two-group discriminant analysis provides a probability of approximately 30 percent of FTA’s and 98 percent for nonFTA’s, thereby indicating that further research needs to concentrate on refining FTA sub classifications, and (3) regression analysis in this study using a dichotomous dependent variable and missing data is comparable to Wilson’s study using a continuous dependent variable and missing data. Further development of the discriminant analysis method appears to be warranted for the following reasons: (1) separate probabilities are derived for both FTA and nonFTA subgroups, (2) discriminant analysis is a statistical method developed specifically for problems dealing with dichotomous dependent variables, and (3) future subclassifications of FTA’s into technical, slow, and fugitive subgroupings should significantly affect probability placements. The paper develops a theoretical framework which conceptually demonstrates the applicability of utilizing an open systems model rather than a closed system model in prediction studies of this nature. Open systems propositions are not derived or chained in the analysis because of the relatively small group placement of 30 percent for the FTA subgroups are numerically enlarged.

Library of Congress Subject Headings

Bail -- Minnesota -- Ramsey County



Number of Pages



South Dakota State University