GEE/Mixed model can i use time as a factor variable? I have data on a compound measured on 1) 0 min, 2) 10min, 3) 60min, 4) 1day and 5) 2months. 
i have coded these time points as 1,2,3,4 and 5, the compound has a cubic function in time when modeled on 1,2,3,4,5. The margins plot looks nice like this but is it appropiate for the data to model it on the factors 1,2,3,4,5? or is it better to use time in minutes/hours/days because of the interpretability, however if this is plotted in a graph the x-axis (time) looks messy, 10 mins-60min-2months is quite disperse?
Thank you 
 A: If you are exploring the data, you may want to actually test this out. Assuming you have some theory about how the compound behaves over time, you could test which specification of time explains or fits the data best. Holding the other covariates/model aspects the same, you could compare the AIC/BIC to the model with time as a continuous variable, time in logs, time as categorical (simply dummies for each time point), and time as a factor or ordinal (as you have already done). Compare the AIC/BIC, and select the best fitting specification.
If, however, in the context of your research you have a strong theory about how the compound behaves over time, then I think you pretty much have to go with that. Fishing around with the time variable to get a better fitting or more significant model is probably not appropriate in a confirmatory setting.
You haven't said if you have other covariates or a treatment condition. If you expect time to depend on these variables, then that may also influence your decision.
