|Date: ||Wed, 29 Jan 1997 21:40:53 -0800|
|Reply-To: ||"L. Shemanski" <lsheman@U.WASHINGTON.EDU>|
|Sender: ||"SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>|
|From: ||"L. Shemanski" <lsheman@U.WASHINGTON.EDU>|
|Organization: ||University of Washington|
|Subject: ||Calibration models using PROC CALIS|
|Content-Type: ||TEXT/PLAIN; charset=US-ASCII|
I have two labs reading the same data and want to calibrate them in a
linear relationship using structural equation models.
If Y is the observed reading from one lab, and X the
observed reading from the other lab (on the same people), the model I want
to use is:
E(Y)=a + b*E(X),
where the models on the "manifest" observed variables are
Y=k + E(Y)*m and X=g + E(X)*h where k, m, h and g are constants, as are
a and b. Normally distributed error variances for all terms, of course.
In order to fully specify this model, I think I need variance and
covariance estimates, which I can get from duplicate readings performed at
each of the labs.
My question, how to perform this analysis in SAS? Any other ideas as to
other possible SAS procedures to use?
Any assistance will be greatly appreciated.