Four-Variable Causal Models and Partial Correlations*

Simon's method yields results that are consistent with those obtained through partial correlations, but it also forces one to make explicit assumptions about both outside disturbing influences and the nature of the various causal links among an entire set of variables. This chapter provides a set of prediction equations for four-variable causal models for the convenience of the nonmathematically inclined social scientist, who may not wish to carry out the computations required by Simon's method. It begins by postulating a causal model involving a given number of variables X_i. The chapter includes additive and linear models, as is commonly done in regression analyses. In nonexperimental situations, such as cross-sectional studies, X_i may not be so plausible as in the case of experimental designs in which randomization has been possible. The introduction of a single additional variable might therefore change a causal relationship from direct to indirect, or even to a noncausal one.