Comment

The article by Conway and Reinganum demonstrates that cross-validation is a necessary part of any goodness-of-fit evaluation of multifactor asset-pricing models. The use of this procedure guards against overfitting, which is a common occurrence in large samples. It is illustrated here that if cross-validation cannot be used, two alternative goodness-of-fit measures that include a penalty based on parameters fitted can be used to assist in model selection. The two alternative measures include a penalty function based on the number of parameters fitted. The values of the two alternative measures for the Conway and Reinganum data are presented here for comparison purposes. The more conservative of the two measures obtains results comparable to the cross-validation procedure.