Date:         Mon, 13 Mar 2000 10:29:36 -0500
Reply-To:     "William B. Ware" <wbware@EMAIL.UNC.EDU>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         "William B. Ware" <wbware@EMAIL.UNC.EDU>
Subject:      Re: Negative Coefficients
Comments: To: Shail Dobson <dobson_shail@HOTMAIL.COM>
In-Reply-To:  <20000313151155.69702.qmail@hotmail.com>
Content-Type: TEXT/PLAIN; charset=US-ASCII

On Mon, 13 Mar 2000, Shail Dobson wrote:

> I did a stepwise regression, 1 dependent variable with 7 independent
> variables. The sample size is 100. It is a rating scale, where 10 means
> excellent and 1 means poor.
>
> One of an important variable shows a positive significant correlation with
> the dependent variables, but in the coefficient matrix this has a negative
> coefficient. It seems funny to me that there would be an inverse
> relationship in the Beta or t-coefficients. Could some shed some light why
> it is happing? I generally conclude my key drivers based on their beta
> values.

Shail,

The result that you describe is often described in the literature as
"suppression."  It is not all that unusual in non-experimental
analyses... The magnitude/pattern of intercorrelation among the
independent variables is the culprit...

For a more complete explanation, see Cohen and Cohen (Applied Regression
Analysis) or Darlington (Linear Models).

WBW

__________________________________________________________________________
William B. Ware, Professor and Chair               Educational Psychology,
CB# 3500                                       Measurement, and Evaluation
University of North Carolina                         PHONE  (919)-962-7848
Chapel Hill, NC      27599-3500                      FAX:   (919)-962-1533
http://www.unc.edu/~wbware/                          EMAIL: wbware@unc.edu
__________________________________________________________________________