The PEAR Method for Sample Sizes in Multiple Linear Regression

Abstract

When multiple linear regression is used to develop prediction models, sample size must be large enough to ensure stable coefficients. If derivation sample sizes are inadequate, the models may not generalize well beyond the current sample. The precision efficacy analysis for regression (PEAR) method uses a cross-validity approach to select sample sizes such that models will predict as well as possible in future samples. The purposes of this study are (a) to verify further the PEAR method for regression sample sizes and (b) to extend the analysis to include an investigation of the effects of multicollinearity on coefficient estimates.