Multiple regression is an extension of the general linear model to include multiple predictors (Allison 1999). The regression of each predictor on the criterion is calculated in a way that only unique variance, separate from the variance between the other predictors and the criterion, is measured. This can be especially useful when there may be correlations among the predictors creating overlapping shares of variance with the criterion. These regression weights are then combined in a linear equation that maximizes the relation of the sum of regression formulae to the criterion. When the linear combination is forced through the origin, standardized beta weights for each of the predictors reflect relative amounts of shared variance.
Multiple regression can be used in an exploratory fashion to find the best prediction of the criterion or in hypothesis testing in which the relation of a specific predictor to the criterion is tested as to whether that predictor...