David A. Kenny
September 6, 2011
Illustration of Deriving Correlations in Terms
of Path Coefficients
Study
����������� Rohner, R. P., Kean, K. J., & Cournoyer, D. E.� (1991).� Effects of corporal punishment, perceived caretaker warmth, and cultural beliefs on the psychological adjustment of children in St. Kitts, West Indies.� Journal of Marriage and the Family, 53, 681-693.�������������������������������
Variables
������������ PP -- physical punishment
���� ������� CR -- caretaker rejection
���� ������� CB -- child beliefs
���� ������� PA -- psychological adjustment
Standardized Structural Equations
������������ CR = .470PP + .883U
������������ CB = .023PP + 1.000V
������������ PA = .752CR + .172PP + .240CB + .610W
������������ rUV = -.477
(Note that because all the variables are standardized, the disturbances have coefficients less than one.)
Tracing Rule
The correlation between Xi and Xj equals the sum of the product of all
the standardized paths obtained from each of the possible tracings between variable
i and j. The set of tracings
includes all the possible routes from Xi
to Xj given that (a) the same variable is
not entered twice and (b) a variable is not entered through an arrowhead and
left through an arrowhead.
������������ rCR,PP = .470
������������ rCB,PP = .023
������������ rCR,CB = (.470)(.023) + (.883)(1.000)(-.477) = -.410
������������ rPA,PP = (.470)(.752) + (.023)(.240) + .172 = .531
������������ rPA,CR = .752 + (.470)(.172) + (.023)(.470)(.240) + (.883)(1.000)(.477)(.240) = .734
������������ rCB,PA = .240 + (.023)(.172) + (.023)(.470)(.752) + (.883)(1.000)(-.477)(.752) = -.065
First Law
Given a standardized structural model, the
correlation between Y and Z where Y is endogenous is equal: rYZ = ΣpYXirXiZ
where p� is
the standardized path or causal parameter from variable Xi to
Y, rXiZ is the correlation between Xi and Z, and the set of Xi variables
are all the causes of the variable Y.
������������ rCR,PP = .470
������������ rCB,PP = .023
������������ rCR,CB = (.470)(.023) + (.883)(-.477)(1.000)� = -.410
������������ rPA,PP = (.470)(.752) + (.023)(.240) + .172 = .531
������������ rPA,CR = .752 + (.172)(.470) + (.240)(-.410) = .734
������������ rCB,PA = .240 + (.752)(-.410) + (.172)(.023) = -.064
The results are the same, although the last one differs a bit due to rounding error.
See Kenny�s 1979 book, Correlation and Causality for more details.
Go
to the Basics page.
Go back to the causal modeling
page.