I am thinking that discriminant analysis could be an option. Can I have some suggestions on other available options?
You may also use logistic regression. I prefer using both logistic regression and discriminant analysis to see whether they provide consistent conclusions.
You did not specify how many categories you have of the dependent variable? If the dependent variable is dichotomous, then you can use probit/logit regression. If the dependent is ordinal, then you should use ordinal logit/probit models, and if the dependent variable is nominal categories, then you can use multiple logit/probit models.
Then, in general, you should go with a logit or probit model if you want to test relationships between the covariates and the outcome. However, if you are more interested in fine-tuning predictions (ie., who is likely to have the outcome = 1 vs outcome vs 0, you should consider more complex models, such as boosted regression, or regression trees.
I hope this helps!
- Badges
- Science topic
- Similar topics
- Philosophy
- Ethics