SPSSX Discussion

CatReg vs GLM

Classic

List

Threaded

3 messages Options

Poling, Taylor Leigh

CatReg vs GLM

Hi all,

I have a relatively small sample size of cases (N = 122) organized into 4 clusters of various size (20, 28, 35, 39). I have a DV with a SD = .60. I used a GLM to examine if cluster type explained variance in my DV. The GLM results were not significant (F = 1.68, df = 3, sig = .17, eta = .20). Then I used the CatReg procedure in spss 17 to examine the same question b/c I had read it uses a bootstrapping method to estimate the standard error thinking the results from this procedure would be less affected by my sample size. So I ran the procedure defining my cluster variable as a nominal variable. These results showed the relationship between cluster and my DV was quite significant (F =13.0, df = 3, std. beta = .25, bootstrap error estimate = .07, sig = .00). My question: Is the difference in these results a product of the bootstrapped error estimate or is there something else I am missing?

Thanks in advance for any insights,
Taylor

Henrik Lolle

Re: CatReg vs GLM

Hi Taylor,

I do not think that the main reason for the small p-value is the
bootstrapping method. In my opinion CATREG have, in many situations,
too liberal significance testing. Have you tried the same model in
CATREG, but without the bootstrapping? The betas from transformed
variables in CATREG is really a good thing, but I think that the
significance testing in many situations should by done with the
original variables in the ordinal ANOVA way.

Best,
Henrik

Quoting "Poling, Taylor Leigh" <[hidden email]>:

> Hi all,
>
> I have a relatively small sample size of cases (N = 122) organized
> into 4 clusters of various size (20, 28, 35, 39). I have a DV with a
> SD = .60. I used a GLM to examine if cluster type explained variance
> in my DV. The GLM results were not significant (F = 1.68, df = 3, sig
> = .17, eta = .20). Then I used the CatReg procedure in spss 17 to
> examine the same question b/c I had read it uses a bootstrapping
> method to estimate the standard error thinking the results from this
> procedure would be less affected by my sample size. So I ran the
> procedure defining my cluster variable as a nominal variable. These
> results showed the relationship between cluster and my DV was quite
> significant (F =13.0, df = 3, std. beta = .25, bootstrap error
> estimate = .07, sig = .00). My question: Is the difference in these
> results a product of the bootstrapped error estimate or is there
> something else I am missing?
>
> Thanks in advance for any insights,
> Taylor
>
>

Kooij, A.J. van der

Re: CatReg vs GLM

In reply to this post by Poling, Taylor Leigh

-----Original Message-----
From: Kooij, A.J. van der
Sent: 12 June 2009 16:14
To: 'Henrik Lolle'
Subject: RE: Re: CatReg vs GLM

The F-test for the coefficients in CATREG is an approximation by analogy
with linear regression and might often indeed be too liberal, so we
recommend to interpret significance results conservatively.

Regards,
Anita van der Kooij
Data Theory Group
Leiden University

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Henrik Lolle
Sent: 12 June 2009 09:05
To: [hidden email]
Subject: Re: CatReg vs GLM

Hi Taylor,

I do not think that the main reason for the small p-value is the
bootstrapping method. In my opinion CATREG have, in many situations, too
liberal significance testing. Have you tried the same model in CATREG,
but without the bootstrapping? The betas from transformed variables in
CATREG is really a good thing, but I think that the significance testing
in many situations should by done with the original variables in the
ordinal ANOVA way.

Best,
Henrik

Quoting "Poling, Taylor Leigh" <[hidden email]>:

************************************************************
Henrik Lolle
Department of Economics, Politics and Public Administration Aalborg
University Fibigerstraede 1 9200 Aalborg
Phone: (+45) 99 40 81 84
************************************************************

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command SIGNOFF SPSSX-L For a list
of commands to manage subscriptions, send the command INFO REFCARD

**********************************************************************
This email and any files transmitted with it are confidential and
intended solely for the use of the individual or entity to whom they
are addressed. If you have received this email in error please notify
the system manager.
**********************************************************************

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD