Abstract
This paper reviews the considerations in evaluating the skill and significance of screening multiple linear regression (SMLR) models. Formulations and procedures are given along with relevant references to prior studies. Topics discussed include predictor selection, serial correlation, artificial skill, true skill, and Monte Carlo significance testing. New results with wide applicability in the assessment of SMLR model skill and significance are presented in graphical form. However, the results are restricted to situations involving predictors which are independent of one another and are serially uncorrelated. The methodology presented is suggested for use in both model evaluation and experimental design.

This publication has 1 reference indexed in Scilit: