LISTSERV at the University of Georgia
Menubar Imagemap
Home Browse Manage Request Manuals Register
Previous messageNext messagePrevious in topicNext in topicPrevious by same authorNext by same authorPrevious page (September 2005)Back to main SPSSX-L pageJoin or leave SPSSX-L (or change settings)ReplyPost a new messageSearchProportional fontNon-proportional font
Date:         Mon, 19 Sep 2005 17:23:38 -0400
Reply-To:     Lisa Stickney <Lts1@enter.net>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Lisa Stickney <Lts1@enter.net>
Subject:      Re: About stepwise Linear Regression
Content-Type: text/plain; format=flowed; charset="iso-8859-1";
              reply-type=original

Hi Karl,

I think it might help if you understand how SPSS selects variables in stepwise regression. In forward stepwise, the first predictor selected is the one that contributes the greatest amt to the model (i.e. highest R-squared). On the second pass, from the remaining predictors, a second predictor is selected using the same criteria, and so on. Stepwise finishes when there are no more predictors that make a significant contribution to the model.

Backwards stepwise works in the opposite direction. It starts with all possible predictors, and one by one removes those that make the least contribution to the model. So, if you run stepwise (in either direction) more than once, you should get the same model. However, that does not mean that forward will produce the same model as backward stepwise or as the original model you tested.

Now the caveat: Just because the variables contribute significantly in some specified order, it doesn't mean that there isn't a better combination of variables that can predict your DV (some combination that doesn't neatly fit the stepwise algorithm). Also, please remember that the model selected should always be grounded in theory, if not, it's data mining (a big taboo in management research).

I hope this helps.

Best, Lisa

Lisa T. Stickney Ph.D. Student The Fox School of Business and Management Temple University

----- Original Message ----- From: "Hector Maletta" <hmaletta@fibertel.com.ar> Newsgroups: bit.listserv.spssx-l To: <SPSSX-L@LISTSERV.UGA.EDU> Sent: Monday, September 19, 2005 12:19 PM Subject: Re: About stepwise Linear Regression

> Stepwise regression produces one model per step. So if you have ten > independent variables, up to 10 models will be produced. If some variables > are not significant, the total number of models may be less. > At each step variables are added and/or taken out of the previous model. > What variables are introduced or taken out at each step depends on the > criteria used (the default criteria or some other values established by > the > user). > Besides, stepwise can be forward or backward, i.e. starting with no > independent variable and then adding one by one, or starting with all the > variables in the equation and proceeding to exclude one by one. One > particularly tricky issue is that while adding variables (forward) or > excluding variables (backward) is the main business of stepwise > regression, > it also does the opposite along the way. Limiting ourselves for clarity to > forward stepwise, at each step SPSS does the following: (a)it includes one > more variable in the previous step's equation, and (b) it MIGHT exclude > some > variable that it had previously included in it. So variable Z may be > entered > into the equation at step 4, remain there at step 5 and 6 while other > variables are introduced, and then be excluded from the equation in step > 7. > SPSS has criteria and thresholds for inclusion and exclusion of variables > at > each step. > > Therefore it is understandable that a variable deemed significant in one > step may end up excluded at a subsequent step. The reason is that > variables > introduced afterwards, even if none has individually the strength of the > variable in question, may add up to the effect that the originally > included > variable turns out to be redundant. > > In backward stepwise the process is the opposite: a variable eliminated > from > the equation at step 3 may end up included again at step 8. > > Hope this helps understanding the process. The question may involve other > issues, but it is not clear to me what else may be involved. > >> -----Original Message----- >> From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] >> On Behalf Of Karl Koch >> Sent: Monday, September 19, 2005 12:55 PM >> To: SPSSX-L@LISTSERV.UGA.EDU >> Subject: About stepwise Linear Regression >> >> Hello list, >> >> When doing stepwise linear regression with SPSS 11.5 I got >> two models. One of the models excludes a former highly >> significant factor. Why is that? What does it in general mean >> if stepwise regression produces more than one model? >> >> Karl >> >> -- >> GMX DSL = Maximale Leistung zum minimalen Preis! >> 2000 MB nur 2,99, Flatrate ab 4,99 Euro/Monat: >> http://www.gmx.net/de/go/dsl >> >> __________ Informacisn de NOD32 1.1220 (20050919) __________ >> >> Este mensaje ha sido analizado con NOD32 Antivirus System >> http://www.nod32.com >> >> >


Back to: Top of message | Previous page | Main SPSSX-L page