I am trying to make predictions of the outcomes of a nominal variable (3 levels) given 4 independent variables that are also nominal. Which approach should I use? If my independent variabbles were continous I was going to use multinomial regression. Is it also applicable here?
Thank you in advance
Yes. Logistic regression is one such if the dependent variable has only 2 levels. If the dependent variable has more than 2 levels, refer to
(1) Albert,A.,A nderson,J.A.(1984).On the existence of maximum llikelihood
estimates in logistic regression models .Biometrika. 71(1):1–10.
(2) Silvapulle,M.J.(1981).On the existence of maximum likelihood estimators for the binomial response models. J. Royal Stat. Soc. Ser.B.43(3):310–313.
(3) G. Zeng (2017). On the existence of maximum likelihood estimates for weighted logistic regression. Communications in Statistics - Theory and Methods. Communications in Statistics - Theory and Methods, 46:22, 11194-11203, DOI: 10.1080/03610926.2016.1260742
Hi:
If the dependent variable is nominal, you should use a multinomial logit model. It does not matter if your independent variables are nominal or continuous.
Thanks
Divyakant
Thank you for the responses, Indeed the multinomial approach can still be applied.
I confirm that the use of multinomial models is a good solution. I highly recommend you to study the relevant chapters in the books:
Long J. S, Freese J. "Regression Models for Categorical Dependent Variables Using Stata", Stata Press, 2006.
Hilbe J. M. "Logistic Regression Models", Chapman & Hall/CRC, 2009.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression