Optimal Scaling and Automatic Linear Regression

Abstract

Optimal scaling is a method designed to optimize the statistical power of the relationship between the predictor and outcome variables. It makes use of processes like discretization (converting continuous variables into discretized values), and regularization (correcting discretized variables for overfitting, otherwise called overdispersion). The current chapter gives examples and shows, that in order to fully benefit from optimal scaling a regularization procedure is important. This chapter also addresses automatic linear regression in SPSS. This provides much better statistics of these data than traditional multiple linear regression does. Optimal scaling is a major contributor to the benefits of the automatic linear regression module in SPSS statistical software. We conclude, that optimal scaling using discretization, is a method for an improved analysis of clinical trials, where the consecutive levels of the variables are unequal. In order to fully benefit from optimal scaling, a regularization procedure for the purpose of correcting overdispersion is desirable.