Date: Wed, 26 Aug 1998 16:51:12 +0200
Reply-To: "Christian F. G. Schendera" <cschende@IX.URZ.UNI-HEIDELBERG.DE>
Sender: "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From: "Christian F. G. Schendera" <cschende@IX.URZ.UNI-HEIDELBERG.DE>
Subject: partial-option in PROC CANCORR
Content-Type: text/plain; charset="utf-7"
For whatever reason, this posting has not found its way into SAS-L. I am
I apologize for any duplicates. I want to calculate the partial correlations
within a multivariate model. SAS/STAT recommends PROC CANCORR. My data
consist of three sets of variables. Set 1 contains moderator variables (all
interval). Set 2 contains
dependent variables (all interval, too). Set 3 contains independent
variables (all nominal, but numerical coded).
First, can I calculate the partial correlations within/between these three
sets with PROC CANCORR at all?
Second, Vars listed in the VAR or WITH-statement do have to be
numerical ones. Can I put set 3 in the PARTIAL-statement, variables
numerical coded but nominal?
Third. The same to the VAR- and WITH-statement. What SAS/STAT Vol.1 does not
say: Can I put vars in the VAR or WITH-statement that are numerical,
but nominal with two or three levels (e.g. TEXT, nominal, length of text
coded as 1, 2, 3)? Tried some variants and didn't find any differences in
standardized results, but may have overlooked something.
Find example procedure below. Sorry to bother you with that, but there is no
one experienced around to ask. I am sure someone out there has already
experience with PROC CANCORR. Have a nice weekend,
Set1: ALS RSA NAAS SKPP AGE AMBINTOL SCHOOL CITY (mod. var, interval)
Set 2: RATING HAS (dep. var, interval)
Set 3: TEXT TREAT NAME (to be partialled out, but: nominal var)
Attention: All vars of set 3 in PARTIAL-statement are ordinal, but
numerical coded. Every other variable of Set 1 and 2 (VAR, WITH) is
numerical and interval
example (most signs are missing. they produce MS-clutter):
vdep vreg wdep wreg
var ALS RSA NAAS SKPP AGE AMBINTOL SCHOOL CITY
with RATING HAS
partial TEXT TREAT NAME