Confound-isolating cross-validation approach to control for a confounding effect in a predictive model.
