On the Interpretation of the Partial Regression Coefficient in Multiple Regression Analysis

Abstract

A lot of textbooks of multiple regression analysis explain how to interpret partial regression coefficients and how to select variables from the viewpoint of predictions, while it is not easy to understand them when the effect of the intervention to some variable is considered, which may give rise to misunderstandings. This paper describes how to interpret the partial regression coefficients by taking account of direct, indirect effects and spurious correlations based on causal diagrams.