Philippe Cattin

A predictive-validity-based procedure for choosing between regression and equal weights

Abstract The equal weighting procedure has often been found to have more predictive validity than regression. This paper makes use of estimators of the population cross-validated correlation to estimate the predictive validity of both regression and equal weights and to suggest a procedure for choosing between the two in any given situation.

Comment on a Monte Carlo investigation of scaling as an alternative to regression

Abstract In a recent study, Emery found that MONANOVA is “superior” to regression as a method of approximating data, whether the measurement scale of the criterion variable is interval or ordinal. However, there are three major reasons for these results. With comparison methods that are fairer to both analytical procedures, the performances of regression and of MONANOVA were found to be close even when the measurement scale of the criterion variable is ordinal.