ABSTRACT
Multiple regression is a statistical technique designed to assess the association between multiple variables and the outcome variable. Regression may be conceived as an extension of correlation, with the goal being prediction of an outcome from a set of predictor variables. One advantage of using regression over correlation is the technique allows the overlaps or correlations across predictor variables to be taken into account when predicting an outcome. Multiple regression can be used to assess the most important predictor variables from a set. Because of preexisting correlations between predictor variables, correlations alone may be inaccurate to assess the importance of variables in their association with an outcome since independent variables will typically overlap. Hierarchical linear regression is preferable for model building because the investigator controls the order of entry of predictor variables based on theoretical considerations as well as less capitalization on chance.
