Data were obtained from 186 applicants at a large Midwestern bank to test the hypothesized advantage of employing a continuous as opposed to a dichotomous criterion when predicting consumer installment credit risk. A continuous criterion of cost was derived and regressed on 21 background items obtained from the application blank (e.g., number of dependents, total monthly income, present indebtedness, etc.). The maximum step of the regression analysis yielded a multiple squared correlation coefficient of .24. This data was then subjected to multiple discriminant analysis employing a median split criterion which yielded an eta squared correlation coefficient of .13. As can be inferred from these results, a greater degree of predictive accuracy may be attained when predicting the continuous variable of risk. Further indicative of the hypothesized advantage of employing a continuous opposed to a dichotomized criterion were the results of the bivariate correlational analyses. Nine background items correlated significantly with the continuous criterion while only five correlated significantly with the dichotomized criterion.
Identifer | oai:union.ndltd.org:WKU/oai:digitalcommons.wku.edu:theses-3655 |
Date | 01 January 1977 |
Creators | Montebello, Anthony |
Publisher | TopSCHOLAR® |
Source Sets | Western Kentucky University Theses |
Detected Language | English |
Type | text |
Format | application/pdf |
Source | Masters Theses & Specialist Projects |
Page generated in 0.0014 seconds