Selecting The 'Best' Prediction Model: An Application To Agricultural Cooperatives

ables suggested by the theory. Stepwise procedures, A credit scoring function incorporating statistical for instance, have been used for variable selection selection criteria was proposed to evaluate the credit in bankruptcy prediction. These procedures, howworthiness of agricultural cooperative loans in the ever, examine variables in a sequence usually deterFifth Farm Credit District. In-sample (1981-1986) mined by the data. The purpose of this study is to and out-of-sample (1988) prediction performance introduce a procedure that supplements previous of the selected models were evaluated using rank studies by giving further consideration to the specitransformation discriminant analysis, logit, and fication of a statistical model within a management probit. Results indicate superior out-of-sample per- oriented framework and the evaluation of predictive formance for the management oriented approach performance of that model. Four specific objectives relative to classification of unacceptable loans, and were associated with the classification and predicpoor performance of the rank transformation in out- tion of agricultural cooperatives into two groups of-sample prediction. (acceptable and unacceptable) according to the performance criteria provided by the Bank for Coopera