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, 12 Sep 2005 09:06:52 -0500
Reply-To:     "Granaas, Michael" <>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         "Granaas, Michael" <>
Subject:      Re: multiple regression model
Comments: To: Omar Farook <>
Content-Type: text/plain; charset="iso-8859-1"


Was this intended as exploratory research? I ask because the sample size is extremely small for any sort of statistically reliable results. You also ask about expanding the subject count and reruning the analysis which suggests this may have been intended as a preliminary study of some sort.

Given the small sample size you really don't have much information as to the quality of this model. If you want to know anything about this model you need more subjects...approximately 20-30 per predictor (140 - 210)at a minimum.

By testing the overall model you are also allowing intercorrelations among the predictors to become partialed out so that you are only testing the unique shared variance between each predictor and the outcome after shared variance with the rest of the predictors has been partialed out. Testing the overall model is largely uninformative.

The best approach is to start by identifying a theoretically meaningful order for entering your predictors into the regression model. Then, after appropriate screening, enter your variables one at a time in the specified order.

If you have no theoretically meaningful way to order your predictors you may wish to simply test each predictor alone to see if it predicts the outcome variable. With an adequate sample size this should at least allow you to reduce the number of predictors considered in future studies.

Screening: Look at scatterplots to make sure that the possible relation between each predictor and the outcome are linear. Do this before testing regressions. It makes little sense to test for linear relations if the data show a clear non-linear relation.

MG **************************************************** Michael Granaas Assoc. Prof. Phone: 605 677 5295 Dept. of Psychology FAX: 605 677 3195 University of South Dakota 414 E. Clark St. Vermillion, SD 57069 *****************************************************

-----Original Message----- From: SPSSX(r) Discussion on behalf of Omar Farook Sent: Fri 9/9/05 11:11 PM To: SPSSX-L@LISTSERV.UGA.EDU Subject: multiple regression model Dear Friends.

I have multiple regression model consist of one dependent and 7 independent variables.

Bu using SPSS I have got the figures below.

t statistic of the 7 predictors variables are -1.728,2.513,-1.580,-0.604,-0.667,2.627 and 1.953.

Sig are 0.159,0.066,0.189,0.578,0.541,0.059 and 0.123.

Sample size =12.

According to my knowledge I should look for a t values below -2 or above +2 and Sig or p-value of less than 0.05 is considered significant.

From the figures above we see that there is no liner relationship between the dependant and an independent variables.

My question is, in this case is it correct to expand the sample size and rerun the same model or I should look for new independent variables?

Hope I was clear enough.

Thanks in advance.


--------------------------------- Click here to donate to the Hurricane Katrina relief effort.

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