topic Re: Correlation Matrix in SAS Procedures

Correlation Matrix

kthartma — Sat, 24 Mar 2018 20:17:46 GMT

Hi there,

I have been using proc corr to conduct correlation matrix for continuous variables: "var" diagnosis rates (for 8 years) "with" volumes (for 8 years) - see below.

I want to do something similar for "var" diagnosis rates (8 years) "with" sex (m/f) and another categorical variable with 5 levels. Any suggestions for procedures to do a correlation matrix (var: 8 continuous, with categorical [2 levels and another variable with 5 levels])?

PROC CORR DATA=data.eva;
VAR rate_2009 rate_2010 rate_2011 rate_2012 rate_2013 rate_2014 rate_2015 rate_2016;
WITH tot_2009 tot_2010 tot_2011 tot_2012 tot_2013 tot_2014 tot_2015 tot_2016;

run;

Re: Correlation Matrix

PeterClemmensen — Sun, 25 Mar 2018 06:42:56 GMT

Do you want to create separate correlation matrices for each level of your categorical variable?

Re: Correlation Matrix

kthartma — Sun, 25 Mar 2018 15:37:15 GMT

Hi,

I think so, I want it to be like:

Tot_2009 Tot_2010 Tot_2011 Tot_2012
Female
Male

If that makes sense. Let me know.

Re: Correlation Matrix

PaigeMiller — Sun, 25 Mar 2018 19:50:03 GMT

@kthartma wrote:

Tot_2009 Tot_2010 Tot_2011 Tot_2012
Female
Male

If that makes sense. Let me know.

Actually, it doesn't make any sense to me. This is just words without description or meaning in the context of correlation of matrices.

The only thing that makes sense to me in the context of correlation matrices is to produce a correlation matrix for males and a different correlation matrix for females. If that's the case, sort by male/female, and then put a BY statement into your PROC CORR.

If that's NOT what you mean, please describe this in detail, instead of the rather brief descriptions you have given so far. Show us an example of what you want.

As long as we're discussing this, I don't see any reason to correlate rate2009 with tot_2016, for example, I can't image how this correlation of data 7 years apart would be meaningful.