A simple index of prediction accuracy in multiple regression analysis

Abstract

Abstract Objectives Within the context of multiple regression the coefficient of determination can be converted to a probability of agreement between the actual and predicted outcomes, suitably dichotomized. Methods This probability of agreement can be used as a simple index of prediction accuracy to help capture the probability of a correct prediction in multiple regression. Results The simple index of prediction accuracy makes the multiple correlation comprehensible to statisticians and laypeople alike. Two examples are provided to demonstrate the application of the simple index. Conclusions In short, the paper introduces the simple index, its computation formula, and its theoretical affinity to the confusion matrix, binomial effect size, probit model, and tetrachoric correlation.