When less is more. Data reduction in the prediction of postoperative outcome

Abstract

In this edition of Epilepsy & Behavior, St-Laurent et al. examine the ways in which test scores from multiple neuropsychological tests can be reduced via principal component analyses into easily interpretable latent components reflecting IQ and verbal and visual memory functions. This approach has considerable intuitive appeal. While some success has previously been reported using individual scores from memory tests to predict cognitive outcomes following epilepsy surgery, potentially valuable data can be lost with the problem of multicollinearity in multiple regression analyses.