topic Re: pearson partial correlation with PROC CORR and PROC GLM in Statistical Procedures

pearson partial correlation with PROC CORR and PROC GLM

Sandrapharma — Thu, 24 Jun 2021 21:11:41 GMT

I am trying to calculate the pearson partial correlation between continuous variables, x and y adjusted by continuous variable z. based on this article https://www.lexjansen.com/pharmasug/2010/SP/SP01.pdf, the following two procedure will give the same results (Pearson correlation from PROC CORR will be the same as the sqrt(R Square) from PROC GLM. However, I always got different results. I am wondering why and which one I should use to calculate the partial correlation.

PROC CORR data = adqol;
var y;
with x;
partial z ;
run;

PROC GLM data=adqol;
model y = x z;
manova / printE;
run;
quit;

Re: pearson partial correlation with PROC CORR and PROC GLM

PaigeMiller — Thu, 24 Jun 2021 21:15:09 GMT

@Sandrapharma wrote:

I am trying to calculate the pearson partial correlation between continuous variables, x and y adjusted by continuous variable z. based on this article https://www.lexjansen.com/pharmasug/2010/SP/SP01.pdf, the following two procedure will give the same results (Pearson correlation from PROC CORR will be the same as the sqrt(R Square) from PROC GLM. However, I always got different results. I am wondering why and which one I should use to calculate the partial correlation.

Show us the results.

Re: pearson partial correlation with PROC CORR and PROC GLM

Rick_SAS — Fri, 25 Jun 2021 10:09:39 GMT

I don't think your PROC GLM step is correct. You need to put both variables on the left hand side of the MODEL statement. Try this:

PROC CORR data = sashelp.class;
var Weight;
with Height;
partial Age;
run;

PROC GLM data=sashelp.class plots=none;
model Weight Height = Age;
manova / printE;
run;
quit;

Re: pearson partial correlation with PROC CORR and PROC GLM

Sandrapharma — Tue, 29 Jun 2021 06:05:09 GMT

thanks. it has been resolved after i put both variables on the left of the model statement for GLM