Currently I am working to defend my thesis, and I ran into a technical question while analyzing my data: Should I conduct an ordinal or a multiple regression?
My model is set up such that I have one Yes/No (dichotomous) predictor, one continuous predictor, and one outcome that is categorical.
Hello Steve Baumgartner. You said you have "one outcome that is categorical."
You might consider ordinal logistic regression if the categories have inherent order or multinomial logistic regression if they do not.
I think the only circumstance in which I would consider multiple (OLS linear) regression would be if the number of categories was fairly large and if I could make a reasonable and defensible claim that the outcome had approximate interval scale properties (e.g., such that means and SDs were sensible and defensible for description). HTH.
Hello Steve Baumgartner. You said you have "one outcome that is categorical."
You might consider ordinal logistic regression if the categories have inherent order or multinomial logistic regression if they do not.
I think the only circumstance in which I would consider multiple (OLS linear) regression would be if the number of categories was fairly large and if I could make a reasonable and defensible claim that the outcome had approximate interval scale properties (e.g., such that means and SDs were sensible and defensible for description). HTH.
Hello, Stephen Baumgartner. Hoping you're having a good day.
If there is any natural order among the categories of Outcome variable, then you should definitely go for Ordinal Regression, Otherwise you can conduct multinomial logistic regression.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression