Hi,
The multinomial regression I am running in spss gives errors, "insufficient memory to complete the model estimation step... execution of this command stops" even though the factor variables are complete and all the covariate variables as well. I have tried to run the variables separately to the dependent variable but only one of the variables works, does anybody have experience with Nom Regressions?
My dependent variable constitutes 2478 nominal categories that are repeated
my independents are
1 years of experience -logged- (scale)
2 industry (ordinal categorical)
Thanks for the suggestions!
Hello Eliaan. First, as this is a question about how to make something work in SPSS, I think you'll have better luck if you post to the SPSSX-L forum (see link below).
Second, when you post there, please clarify what you mean when you say the "dependent variable constitutes 2478 nominal categories that are repeated". On the face of it, it sounds like the outcome variable has 2478 distinct (nominal) categories, and (possibly) that you have repeated measurements of the DV for each subject. If that is indeed what you mean, having that may outcome categories will cause problems. Multinomial logistic regression is generally used with a much smaller number of outcome categories (e.g., 3-7 or so).
It would also help if you indicate which procedure you are using (is it NOMREG?), and if you posted your syntax and the exact error message. Also report your sample size, and the frequency counts for the outcome variable categories.
HTH.
http://spssx-discussion.1045642.n5.nabble.com/
Hi,
I'm not completely sure I understand your model, but you may have at least two problems. First, if you have repeated measures you violate the assumption of independence using any kind of linear regression. There are ways to deal with repeated measures in a generalized linear regression setting, but it may be worth thinking through whether time is part of your model. However, I doubt non-independence would throw an error.
The bigger problem is if you really have something like 2500 categories in your dependent variable. That's too many. Effectively, you're asking your software to compute 2477 equations with two independent variables. I don't have a handy reference for the maximum number of categories, I'd guess you'd want it to be under ten for sure, and five or fewer most likely. Can't you (empirically or pragmatically) collapse a couple thousand of those?
You don't say what your sample size is, but small sample sizes won't permit complex models to run.
I may have misunderstood your problem, but I hope this helps.
A good reference for categorical data analysis is Agresti:
Agresti, A. (2002). Categorical Data Analysis. 2nd. ed. Wiley.
Here is a free course on nominal outcome models
https://www.cmm.bris.ac.uk/lemma/
and more specifically
https://www.cmm.bris.ac.uk/lemma/course/view.php?id=13#section-11
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS