This short paper tested the value of equal weights versus regression weights in predictive models. It goes back to the problem posed in the influential paper by Dawes and Corrigan (1974); that is, how can one predict the success of graduate students? Remus, using a validation sample of 100 University of Hawaii MBA students, obtained the same result as had Dawes and Corrigan; the regression coefficients did not add predictive ability. In fact, equal weights correlated a bit higher to success than did the model with regression weights (0.139 versus 0.132). However, Remus argues that a more relevant criteria is to look at the percentage of students who were successful in the program. Here, regression weights were 74% correct versus 57% for equal weights. Reference
