Multinomial regression in SPSS
Locked
Multinomial regression in SPSS
 
|
Hello
This is a query about running (unordered) Multinomial logistic regression in SPSS. 1. If I run the model using the Analyze>Regression>Multinomial logit option, I'm don't see an option for changing the reference category for any factor/categorical variables, so that the last category is always used as the reference group. I would therefore use this group as the reference group or recode the categorical variable accordingly. Is that correct? 2. Using the Generalized Linear Models Option, if I custom build a multinomial model using the cumulative logit link function, I end up with an ordered logistic model, where as I want an unordered multinomial model. Can this be built using the GLM menu? Many thanks Ron |
Re: Multinomial regression in SPSS
|
|
Embedded. Gene Maguin From: SPSSX(r) Discussion [mailto:[hidden email]]
On Behalf Of Ronald McDowell Hello >> yes.
>> What happens if you use logit link?
|
Re: Multinomial regression in SPSS
|
|
In reply to this post by Ronald McDowell
Hi Ron, I think you may be correct on both cases. 1.
For the multinomial procedure: I did not see the option for changing the default option of reference category for the factors, either. The option
is not available under the syntax environment as well as the point-and-click environment. I would worry about the coding of the data file, instead of the coding of the procedure. 2.
Under the GENLIN procedure, it does only the ordinal logistic regression (using the cumulative logits link, instead of the generalized logits link
for nominal categorical data) If I am correct, the genlinmixed procedure does the nominal logistic regression (Mixed Models/Generalized Linear), if the scale of measurement of the DV under
the Variable View is set up properly. The distribution for DV should be multinomial and the link function should be logit. Then, if the scale of measurement of the DV is specified as Nominal, the procedure will use the generalized logit link. If the scale
of measurement of the DV is specified as Ordinal, the procedure will use the cumulative logit link. And in this procedure, you are able to change the reference level of the factors. Hongwei “Patrick” Yang, University of Kentucky From: SPSSX(r) Discussion [mailto:[hidden email]]
On Behalf Of Ronald McDowell Hello |
Re: Multinomial regression in SPSS
|
|
In reply to this post by Maguin, Eugene
GENLIN (generalized linear models) only
allows for an ordinal multinomial dependent, and not a nominal multinomial
dependent. GENLINMIXED (generalized linear mixed models) allows for
either an ordinal or nominal multinomial dependent.
Alex From: "Maguin, Eugene" <[hidden email]> To: [hidden email], Date: 08/19/2013 10:54 AM Subject: Re: Multinomial regression in SPSS Sent by: "SPSSX(r) Discussion" <[hidden email]> Embedded. Gene Maguin From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Ronald McDowell Sent: Monday, August 19, 2013 4:14 AM To: [hidden email] Subject: Multinomial regression in SPSS Hello This is a query about running (unordered) Multinomial logistic regression in SPSS. 1. If I run the model using the Analyze>Regression>Multinomial logit option, I'm don't see an option for changing the reference category for any factor/categorical variables, so that the last category is always used as the reference group. I would therefore use this group as the reference group or recode the categorical variable accordingly. Is that correct? >> yes. 2. Using the Generalized Linear Models Option, if I custom build a multinomial model using the cumulative logit link function, I end up with an ordered logistic model, where as I want an unordered multinomial model. Can this be built using the GLM menu? >> What happens if you use logit link? Many thanks Ron |
| Free forum by Nabble | Edit this page |