Towards an Algebra for Analyzing Causal Relations.
Frederick S. Ellett, David P. Ericson · 1982
A major part of the paper investigates the relationships among correlation, partial correlation, and various conceptions of-causation. ) Two widely used rul --of causal inference are explicated. It is argued that for' each of the onceptions of causation the causal inference rules, which usp 'par ial correlations, are invalid. Another part of the paper explicates the basic principles of path analysis and structural equation analysis. It is shown that these approaches are subject to three different but important problems. The final, major part tentatively develops an alternative approach (the conditional probability approach) which use5 conditional probability and not correlation as the key concept. It is shown that the C.P. approach can avoid the shortcomings and prOblems of such approaches as causal modeling and path analysis. It is also shown that it provides plausible composition and decomposition rules as well as a plausible measure of causal strength. In a supplemental section, we present the theorems which hold for dichotomous systems under two sets of assumptions and the theorems which' hold for continuous systems under two sets of assumptions, theorems where the probabilistic conception of causation is employed. Towards an Algebra for Analyzing,Causal Relations