ABSTRACT
Prediction equations that contain several predictor variables require closer examination and more skills from the data analyst than do single-variable ones. The methodology for single-variable analysis is generalizable to multiple-variable analyses, only in the latter situation there are more interrelationships among variables to consider. Assessing the information contained in these interrelationships necessitates some basic knowledge of matrix algebra and a large amount of insight on the part of the investigator. The need to be concerned about interrelationships among the predictor variables in multiple-variable models is well-illustrated by a controversy described in Blalock. Matrix algebra provides short-cut techniques for expressing multiple-variable prediction equations and the associated computational manipulations needed for a regression analysis. Using matrix notation, multiple linear regression analysis appears as a natural extension of single-variable regression. Blalock points out that the different conclusions can be attributed to either the existence of or the lack of an association between the predictor variables, anomie (AN) and authoritarianism (AU).
