Date: Tue, 22 Jan 2002 09:53:06 -0500
Reply-To: Jay Weedon <jweedon@EARTHLINK.NET>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Jay Weedon <jweedon@EARTHLINK.NET>
Subject: Re: testing the differences of parameter estimates from two data
Content-Type: text/plain; charset=us-ascii
On 20 Jan 2002 19:15:45 -0800, firstname.lastname@example.org (Yvette) wrote:
>I want to test the difference of my parameter estimates from two data
>sets. I know that the wald statistic is a test of the parameter
>estimates however I am not sure how to use this test or a more
>appropriate test across different data sets or across multiple
>regressions within one data set.
I call the predictor variable of interest X. I assume that the model
you've fitted to the two datasets to be identical. (If it's not you
have a difficult problem.)
1. Concatenate the two datasets.
2. Create a dummy variable XSET, corresponding to which of the two
sets each observation belongs to.
3. Create a variable XNEW, the product of XSET & X.
4. Repeat the original analysis on ALL cases, adding XNEW to the list
of predictors. Probably you'll want to add XSET as a predictor as
well, depending on exactly what model you want to fit. If the Wald
statistic for XNEW's parameter indicates it to be significantly
differently from zero, you can conclude that X's parameter differs
between the two sets.