Out-of-bag prediction error: A cross validation index for generalized structured component analysis

This study introduces out-of-bag prediction error (OPE) for evaluating the predictive generalizability of GSCA models. Repeated bootstrap samples provide separate observations for model estimation and prediction assessment, allowing researchers to compare candidate models without collecting a separate test sample. The procedure fits naturally into GSCA’s existing bootstrap workflow. Simulated and empirical data are used to evaluate the index.