Abstract
A stepwise procedure for fitting logit models with categorical predictors is present ed. The method provides for detecting nonrelevant predictor variables and collaps ing the table over them. The procedure is applied to data presented by van Alstyne and Gottfredson in a previous issue. We show that the previously reported dis crepancies between the models emerging from their construction and validation samples can be removed by collapsing the table over the nonrelevant variable age. The results emphasize the importance of properly screening the set of possible pre dictor variables and detecting the relevant ones. The method presented here can help the investigator in this task.
Get full access to this article
View all access options for this article.