Date: Wed, 21 Jan 2004 10:49:44 0800
ReplyTo: John Fava <jmfava@MSN.COM>
Sender: "SAS(r) Discussion" <SASL@LISTSERV.UGA.EDU>
From: John Fava <jmfava@MSN.COM>
Organization: http://groups.google.com
Subject: PROC GLM's "Class" statement (vs. using dummy coding for qual
variables)
ContentType: text/plain; charset=ISO88591
Hello,
I'm having difficulty understanding how or when the "Class" statement
in PROC GLM (for ANOVA) is used versus simply "dummycoding"
categorical variables for use in the model (also in PROC GLM). Is the
"class" statement
simply a easier way of dummy coding categorical data, or is there more
to this?
Both methods appear to "work," but the output is definitely not
comparable. What is the difference?
The data set that I'm working with includes a "test" group, with a
"control" group matched up based on several characteristics (four
other factors, all categorical, including sex, gender, age and marital
status). The dependent variable is continuous. I should also say
that this is not a true experimental design ... the control group was
created AFTER the experiment was conducted.
If you can answer the "CLASS" issue, I'd also like to understand
exactly what's calculated to come up with the LSMEANS.
Thanks!
John Fava
