Date: Wed, 16 Feb 2011 20:55:58 -0500
Reply-To: William Shakespeare <shakespeare_1040@HOTMAIL.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: William Shakespeare <shakespeare_1040@HOTMAIL.COM>
Subject: Imputing categorical variables
I've been reading about imputation of categorical variables. The
sentiment seems to be to perform the imputation as a continuous variable
and not round off. Say I want to run proc glm prior to proc mianalyze.
I'm wondering how to subsequently analyze the data. Clearly, one could
end up with categories that have hundreds or thousands of values. Do I
use a covariance or correlation/means matrix as my input data set? Will
glm accept such input?