Four-Variable Causal Models and Partial Correlations

Simon's method for making causal inferences from correlational data is applied to the various possible four-variable causal models. Prediction equations are given for forty-one models, so that the goodness of fit of any particular model can easily be evaluated without the use of tedious computations. Certain suggestion are also made for handling problems involving a larger number of variables. The major purpose of the paper, however, is to investigate what happens to the correlations between two variables when controls are made for variables which are causally related to these variables in different ways.