ABSTRACT
Multiple regression is ideally suited to social science research because it allows researchers to adjust for confounding variables when experimental designs are not possible. But, such confounders are not always measured. When they are not, researchers can use their understanding of multiple regression to anticipate the direction of bias due to their omission. Multiple regression can also be used to determine whether the effect of a variable on the outcome operates through the mechanisms laid out in theories, by adding measures for those mechanisms to the model. But, care must be taken to interpret these mechanisms, because study designs often do not allow the direction of effects to be unambiguously determined.
