- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Posted 11-22-2013 05:00 PM
(1042 views)
Hello,
I am trying to evaluate the correlation between two independent populations via PROC CORR and Fisher's z test. Is it possible to evaluate these two correlations with proc corr (or some other procedure) without inputting an entire dataset and using just the coefficients and sample sizes?
Thank you.
2 REPLIES 2
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
If you have a variable that represents the count of the number times a record should be repeated for your analysis variable appears then use that in a Freq statement;
Proc corr data=yourdata;
var analysisvar;
Freq CountVariableName;
run;
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
A data step solution would look like:
data have;
input r n;
datalines;
0.1 20
0.4 33
-0.15 108
;
data want;
set have;
fishz=artanh(r)*sqrt(n-3);
prob=0.5*(1-cdf('NORMAL', abs(fishz)): /* Two tailed test */
run;
Steve Denham