Hierarchical Multinominal logistic -Can it be done in spss

[msherman]

Dear list: I am attempting to conduct a hierarchical multinominal logistic regression but when I use the menu there are no selections that allow me to enter particular variables as different stages. Within MRA one has the capability of ordering variables according to one’s theory, etc. One can use Enter on the first step, and the enter on the NeXT step, but within multinominal they are not to be found.  Can this be done via writing syntax? thanks.

 

Martin F. Sherman, Ph.D.

Professor of Psychology

Director of  Masters Education in Psychology: Thesis Track

 

Loyola University Maryland

Department of Psychology

222 B Beatty Hall

4501 North Charles Street

Baltimore, MD 21210

 

410-617-2417

[hidden email]

 

[Bruce Weaver]

I've just been looking at NOMREG in the CSR Manual (aka the "fine" manual), and it would seem you are correct.  Re the /MODEL sub-command, the manual says, "If more than one MODEL subcommand is specified, only the last one is in effect."  It appears one can perform various flavours of stepwise selection (forward, backward, BSTEP and FSTEP), but not hierarchical regression.  

So, what you'll have to do is run NOMREG multiple times.  (One has to do this with procedures like MIXED too, by the way.)  One thing you have to watch for is that as you add more variables, the number of cases could be reduced if there are missing values.  You don't want that, because the likelihood ratio tests for comparing one model to the next are only valid when all models use the same cases.  You could set a filter at the outset for cases that have valid data on all of the variables in the most complex model.  Alternatively, I think you can probably list all explanatory variables for every model (following the BY and WITH keywords), but only include subsets of those variables as needed on the /MODEL subcommand.  

HTH.

msherman wrote

Dear list: I am attempting to conduct a hierarchical multinominal logistic regression but when I use the menu there are no selections that allow me to enter particular variables as different stages. Within MRA one has the capability of ordering variables according to one's theory, etc. One can use Enter on the first step, and the enter on the NeXT step, but within multinominal they are not to be found.  Can this be done via writing syntax? thanks.

Martin F. Sherman, Ph.D.
Professor of Psychology
Director of  Masters Education in Psychology: Thesis Track

Loyola University Maryland
Department of Psychology
222 B Beatty Hall
4501 North Charles Street
Baltimore, MD 21210

410-617-2417
[hidden email]<mailto:[hidden email]>

[msherman]

Bruce:  Thanks, Yes, it looks like I will have to do something like the following.

I can enter all vars in a block, then back out the NS ones (one at a time if I want to be thorough).  Then enter the significant ones from that block along with all the vars from the second block, and back out the NS from the second block (not the first block).  Then repeat the process with each successive block.  A bit tedious, but it is essentially a manual version of backward removal within each block.

Also, see email to Ryan which explains my model in more detail.

Ryan: It is not really stepwise. My steps following a formal process.  Pre-evacuation characteristics: Demos then pre-existing conditions(psych disability and general disability), pre-existing evacuation training,  then evacuation context variables, then evacuation process variables, then evacuation challenges, and finally evacuation outcomes-injuries, loss of coworker.  This is a hierarchical model checking to see if some of the variables are direct or mediated.  In particular whether preparedness training is a mediator.  mfs

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Bruce Weaver
Sent: Tuesday, August 16, 2011 10:15 PM
To: [hidden email]
Subject: Re: Hierarchical Multinominal logistic -Can it be done in spss

I've just been looking at NOMREG in the CSR Manual (aka the "fine" manual), and it would seem you are correct.  Re the /MODEL sub-command, the manual says, "If more than one MODEL subcommand is specified, only the last one is in effect."  It appears one can perform various flavours of stepwise selection (forward, backward, BSTEP and FSTEP), but *not* hierarchical regression.

So, what you'll have to do is run NOMREG multiple times.  (One has to do this with procedures like MIXED too, by the way.)  One thing you have to watch for is that as you add more variables, the number of cases could be reduced if there are missing values.  You don't want that, because the likelihood ratio tests for comparing one model to the next are only valid when all models use the same cases.  You could set a filter at the outset for cases that have valid data on all of the variables in the most complex model.  Alternatively, I think you can probably list all explanatory variables for every model (following the BY and WITH keywords), but only include subsets of those variables as needed on the /MODEL subcommand.

HTH.



msherman wrote:

>
> Dear list: I am attempting to conduct a hierarchical multinominal
> logistic regression but when I use the menu there are no selections
> that allow me to enter particular variables as different stages.
> Within MRA one has the capability of ordering variables according to
> one's theory, etc. One can use Enter on the first step, and the enter
> on the NeXT step, but within multinominal they are not to be found.
> Can this be done via writing syntax? thanks.
>
> Martin F. Sherman, Ph.D.
> Professor of Psychology
> Director of  Masters Education in Psychology: Thesis Track
>
> Loyola University Maryland
> Department of Psychology
> 222 B Beatty Hall
> 4501 North Charles Street
> Baltimore, MD 21210
>
> 410-617-2417
> [hidden email]&lt;mailto:[hidden email]&gt;
>

-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Hierarchical-Multinominal-logistic-Can-it-be-done-in-spss-tp4706306p4706503.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
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

=====================
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