Computing partial (conditional) correlation

[pegahtv]

I wanted to compute the conditional pearson correlation (CPC). First, I wanted to make sure if it is OK to use the absolute pearson correlations in the computation of CPC. Becuase I think only the strength of pairwise correlations should be considered in computing CPC, and not the sign. Second, I am supposing that the more the CPC negative, the less the dependency of the two variables on the third variable and vs.. Is this assumption correct, or should I only look at absolute value of CPC?

thanks in advance for your answer.

 

[lippyka]

Yes, it is ok to use absolute pearson correlations in the computation of CPC. The strength of the pairwise correlations should be considered, but the sign can be ignored. Your assumption is correct - if the CPC is negative, then the two variables are less dependent on the third variable, and vice versa.