Date:         Thu, 3 Oct 2002 10:24:25 -0500
Reply-To:     "Nichols, David" <nichols@spss.com>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         "Nichols, David" <nichols@spss.com>
Subject:      Re: Christian RE: ordered Probit or ordinal regression
Content-Type: text/plain; charset="us-ascii"

SPSS can fit an ordered probit response model, beginning with Release 10, in the PLUM procedure, which is in the Advanced Models module. In the menus, select Analyze->Regression->Ordinal, and in the Options dialog box, choose Probit from the Link dropdown.

David Nichols
Principal Support Statistician and
Manager of Statistical Support
SPSS Technical Support

Got a question for SPSS Technical Support?
Try the AnswerNet at http://www.spss.com/tech/answer

-----Original Message-----
From: Bautista Christian [mailto:cbautista@namrid.sld.pe]
Sent: Wednesday, October 02, 2002 2:02 PM
To: SPSSX-L@LISTSERV.UGA.EDU
Subject: [SPSSX-L] Christian RE: ordered Probit or ordinal regression


Hi Margaret,

I think that SPSS cannot do ordered Probit analysis. SPSS can do only one
model for ordina Logistic regression but there are six ordinal regression
models.

If you want to report OR (odds ratio) you have to use regression models, if
not probit can be applied.

If you decide to apply logistic regression remember that these are these
regression models for ordinal responses:

1. Cumulative logit model (proportions odds)
2. Continuation-ratio model
3. Partial-proportional odds model
4. Polytomous logistic model
5. Adjacent-category logistic model
6. Stereotype logistic model

Proportions odds or cumulative logit model is the most popular ordinal
logistic model used and this is incorporated in SPSS, but you need to check
the proportional odds assumption and this may be checked with the Score
test. If this is violated you need to use partial-proportional odds model.

I have done similar analyses for one paper before using ordinal logistic
regression.

Email me, if you think is neccesary.

Good luck,

/Christian
...........................................................................
M.Christian Bautista Tejeda, MSc.
Biostatistician - GIS Data analyst
NAVAL MEDICAL RESEARCH CENTER (US.NMRCD)
American Embassy
Email:  cbautista@namrid.sld.pe



> ----------
> From:         Ferguson, Margaret R.[SMTP:mferguso@iupui.edu]
> Reply To:     Ferguson, Margaret R.
> Sent:         Wednesday, October 02, 2002 11:17 AM
> To:   SPSSX-L@LISTSERV.UGA.EDU
> Subject:      ordered Probit or ordinal regression
>
> We are working on a paper where our dependent variable  is ordinal (4
> categories, values=none, slight, moderate, high).  We have been advised
> to used ordered Probit analysis.  However, we are not sure about if/how
> SPSS will do this.  Will it?  If not, what is the difference between
> ordered Probit and the ordinal regression function in SPSS regression
> listing?  Will the results be essentially the same?
>
> Thanks in advance for your advice.
>
> ***************************************
> Margaret R. Ferguson
> Assistant Professor
> Department of Political Science
> Indiana University at Indianapolis
> Indianapolis, IN 46202
> phone 317.274.4996
> fax 317.278.3280
> ***************************************
>