Fixed effects models are a great choice for longitudinal #epi research if: 1) your independent variable of interest is time-varying (& variation is present in the data), 2) you have at least 2 dependent variable measurements, & 3) you have an unmeasured time-invariant confounder.
#Epidemiologists love to categorize variables ( #logisticregression). However, sometimes it makes sense to categorize a dependent variable into more than 2 categories (e.g., BMI categories).
Multinomial fixed effects models are a little uncommon in epidemiological research. However, if you are interested in running one: femlogit is a great STATA command for fixed effects models with a multinomial dependent variable ( https://bit.ly/2XJ4y25 ).
I have yet to find an easy to use SAS command or know of a similar command in R. If you do, please share! #epitwitter #econtwitter #statstwitter #rstats