Hi,
I checked the distribution of the outcome for sub challenge (SC) 3-the log form of symptomatic score, and it seems that it's not really a continuous outcome. The frequency of some value is over 15% and that definitely forms a probability mass for discrete values. When I transform it back to the categorical score, the histogram shows a pattern very close to Poisson distribution.
Can we treat the outcome of SC3 as categorical by converting it back to the maximum symptomatic score?
Created by Qian Li qli4 You can model the outcome in any way you chose. Just be sure to provide a numerical prediction on the log-scale. The symptomatic score is an integer, so a log of that is discrete. You can recover the integer symptomatic score by exponentiating, and I don't see why you couldn't treat it as categorical if you want.