Given variables $X, Y, Z_i, i\in{1,...,n}$. I want to calculate the partial correlation

$Cor(X,Y|Z_1, ..., Z_n)$

from the simple correlations $Cor(X,Y), Cor(X,Z_1),$ etc.

I have found a formula here that describes how to obtain the first-order partial correlation from zero-order partial correlations, but I have not found a generalization to higher orders.

So my question: Is there a formula to compute the $n$-th partial correlation from the $(n-1)$-th order partial correlations? And if it exists, is there a proof for it?

  • 1
    $\begingroup$ It depends on how much your "etc" comprises, because I believe you need all the pairwise correlations, including among pairs of the $Z_i,$ because you are essentially regressing $X$ and $Y$ separately against the $Z_i.$ See stats.stackexchange.com/a/108862/919 for the relationship between the multiple regression coefficients and the covariance matrix. (The covariance matrix for the standardized variables is the correlation matrix.) $\endgroup$ – whuber Oct 2 '18 at 14:45
  • $\begingroup$ @whuber Yes, "etc" are all pairwise correlations and thanks, this link is very helpful. I will try to obtain the formula I need using the mentioned relationship $\endgroup$ – Bluescreen Oct 4 '18 at 9:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.