|Date: ||Fri, 7 Feb 1997 08:42:32 -0600|
|Reply-To: ||"Nichols, David" <nichols@SPSS.COM>|
|Sender: ||"SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>|
|From: ||"Nichols, David" <nichols@SPSS.COM>|
|Subject: ||Re: Linearity & multicolinearity|
>From: William J. Feuer[SMTP:wfeuer@NEWSSUN.MED.MIAMI.EDU]
>Sent: Thursday, February 06, 1997 8:52 AM
>To: Multiple recipients of list SPSSX-L
>Subject: Re: Linearity & multicolinearity
>> In < Compare Means > there is an option for linearity tests. What test
>> this & how is to be interpreted? Also, other than a simple correlation,
>> what are the options for verifying whether the assumption of linearity is
>> met (e.g., regression)?
>This test, assuming the group codes for which you're getting the means are
>sensibly ordered, is a test for linear trend in the means (from lowest to
>highest group) and deviation from linear trend, as per Draper and Smith,
>chapter 1. It's a very useful thing that the regression procedure doesn't
>> In < Regression > what test is used to determine colinearity (or
>> multicolinearity)? Also, how do I interpret the output of this test?
>> colinearity statistics, the output shows "tolerance" & "VIF". Any
>> explanations would be appreciated.
>You would like to have the Variance Inflation Factors less than 10. I
>beleive the tolerance is just the reciprocal of the VIF (If I'm wrong,
>someone will say something if I'm wrong).
You're right about the VIF. A minor clarification on the first comment:
REGRESSION doesn't have an automatic facility for testing what is tested
in MEANS and ONEWAY, but it can certainly be made to do it.
Senior Support Statistician