|Date: ||Tue, 19 Feb 2002 15:56:49 -0500|
|Reply-To: ||"Stanfield, Robert" <email@example.com>|
|Sender: ||"SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>|
|From: ||"Stanfield, Robert" <firstname.lastname@example.org>|
|Subject: ||Statistics Question about covariates|
|Content-Type: ||text/plain; charset="iso-8859-1"|
This is a general stats question -- not an SPSS one.
I'm trying to generate a regression equation that will allow me to predict
future values based on several years of data that I have. Many of my
predictor variables vary according to the month and year they originated in
(like interest rates, for example). This creates several alternative
hypotheses for my data -- not a good thing. So to control for this I
included a "static pool" variable that codes for year&month as a covariate.
This works very well in terms of explaining the data set that I have.
The problem is I'm not sure what to do with the variable in terms of the
predictive equation, because future values of the IVs are not necessarily
related to the past values. In other words, assuming an average Static Pool
value or even choosing a future value makes no theoretical sense.
Is it valid, therefore, to simply set that value equal to zero in the
regression equation? What effect, if any, would that have on the
interpretation of future values?
Thanks for any help,