CROSS-VALIDATION IN STEPWISE REGRESSION

In stepwise regression procedures, the method of cross-validatory choice is used to select appropriate cutoff values, F(in) and F(out), which are then used for determining the predictor variables from the full data set to be used in a linear prediction equation. Furthermore, we propose a sequential detection procedure based on the application of cross-validation in stepwise regression to detect outliers and influential observations. By analysing some previously analysed data sets, we find that the proposed procedure performs well and shows much useful information about the unusual structure of the data.