Target/Outcome Data for Categorical Family w/Logit Link in BRMS


#1

Hi, When doing binary logistic regression I recode my categories to 0 and 1 and things run smoothly.

When doing a categorical family model with the logit link, a la a multinomial regression, does BRMS expect what R would consider a nominal, factor data type like “red”, “blue”, “green” or should I recode my outcome variable to something like 1,2,3?


#2

I believe there are multiple posts here on discourse discussing multinomial models with brms.


#3

Thank you. I tweaked my question to something more granular above. Plz see above


#4

Both should work, but I would recommend using a factor variable.