Date:         Fri, 2 Jun 2006 08:32:50 +1000
Reply-To:     Stefano Occhipinti <s.occhipinti@griffith.edu.au>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Stefano Occhipinti <s.occhipinti@griffith.edu.au>
Subject:      Re: Comparing regression coefficients
In-Reply-To:  <8AE883D86DBAAF4E9924E984A06927F20286F1E6@mcintire6.comm.virginia.edu>
Content-Type: text/plain; charset="US-ASCII"

Tony,

Cohen, Cohen, West and Aiken (2003) regression text (Something like
Multiple Regression/Correlation for the Behavioral Sciences--my copy is at
work and I think of it as "CCWA") has exactly the test you are talking
about (assuming these are independent samples, not panel data of some
sort). Also, a quick fix is the econometrical approach (the "Chow test" if
I am not mistaken) where you combine the data from the two samples/years
and include a dummy for sample. You then fit a regression model with the
"substantive predictor of interest", the year/sample dummy and the
dummy/predictor cross-product. The test of the cross-product is the test
of the difference in substantive predictor regression weights across the
samples. In fact, I can't imagine why it wouldn't be algebraically
equivalent to the methods that derive a standard error for the actual
difference in regression weights but I am sure someone with the requisite
background will chime in if need be. Naturally, for a bunch of predictors,
this begins to get unwieldy, YMMV.

HTH,
SO
------------------------------------------------------------
Dr Stefano Occhipinti

Griffith Psychological Health Research Centre
and School of Psychology
Griffith University (Mt Gravatt Campus)

Postal Address:
School of Psychology
Griffith University
Nathan QLD 4111
AUSTRALIA

tel: +617 3735 3372
fax: +617 3735 3388
net: S.Occhipinti@griffith.edu.au




"Baglioni, Tony" <ajb2t@comm.virginia.edu>
Sent by: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
01/06/2006 11:32 PM
Please respond to
"Baglioni, Tony" <ajb2t@comm.virginia.edu>


To
SPSSX-L@LISTSERV.UGA.EDU
cc

Subject
[SPSSX-L] Comparing regression coefficients






All,



I've been searching stat texts for this but I cannot find the answer.  I
have comparable data for two years.  I have obtained the regression
equations for each year and now I want to compare the obtained
coefficients.  I had a text that showed a simple t or z test for doing
this it's been lost through the ages.  Does anyone know a simple test
for this comparison?



Thanks,

Tony



A. J. Baglioni, jr.

Assistant Professor

McIntire School of Commerce

University of Virginia

434 924-4961