topic I am doing a Proc Corr But I only it to output correlation with with correlations higher than 0.7 in SAS Studio

I am doing a Proc Corr But I only it to output correlation with with correlations higher than 0.7

Avdol — Mon, 09 Dec 2019 21:36:16 GMT

I am doing a Proc Corr on a data set But I only want to output the correlations of variables that have absolute correlations higher than 0.7.

I am using SAS Studio this is my code.

ods noproctitle;
ods graphics / imagemap=on;

proc corr data=_TEMP0.BODYFAT2 pearson nosimple noprob rank outp=work.Corr_stats12;

var Age Weight Height Neck Chest Abdomen Hip Thigh Knee Ankle Biceps Forearm Wrist;
with Age Weight Height Neck Chest Abdomen Hip Thigh Knee Ankle Biceps Forearm Wrist;
run;

I tried using the best and rank statement but they aren't really helping me.

Thank you in advance

Re: I am doing a Proc Corr But I only it to output correlation with with correlations higher than 0.

PaigeMiller — Mon, 09 Dec 2019 21:42:24 GMT

You could use the BEST= option to get a data set with the top correlations, and then delete any correlations that are <0.7

Re: I am doing a Proc Corr But I only it to output correlation with with correlations higher than 0.

Avdol — Mon, 09 Dec 2019 22:12:19 GMT

I have tried using the rank function but how do I delete the correlations?

Re: I am doing a Proc Corr But I only it to output correlation with with correlations higher than 0.

FreelanceReinh — Mon, 09 Dec 2019 22:34:56 GMT

Hello @Avdol and welcome to the SAS Support Communities!

If your primary focus is on the printed output (as this is what the options BEST= and RANK only address), you can define and use a numeric format to suppress any low correlation coefficients:

proc format;
value corr07f
-0.7-0.7 = ' '
other    = [8.5];
run;

ods select none;
ods output pearsoncorr=pc;
proc corr data=_TEMP0.BODYFAT2 noprob;
var Age Weight Height Neck Chest Abdomen Hip Thigh Knee Ankle Biceps Forearm Wrist;
run;
ods select all;

proc print data=pc noobs;
format _numeric_ corr07f.;
run;

Maybe you would like to suppress the trivial "1.00000" values on the main diagonal of the correlation matrix as well (provided that you can rule out [almost] perfect correlations elsewhere!). Then just include 1 in the range specification of the format definition:

-0.7-0.7, 1 = ' '

(Edit: simplified format definition.)