Hello everyone,

I am getting started and was wondering what would be the best way to approach categorical variables using stcrreg and stcox for univariable analysis, with the goal of eventually including relevant variables into a multivariable analysis model.

I want to look at the relationship between level of invasion of a tumour and poor outcomes. I have a variable called 'invasion' that has three possible values: superficial, middle, deep. I have also made three separate indicator variables ('invasion_sup', 'invasion_middle', 'invasion_deep') that store either yes or no. Which one of the following approaches should I be using for univariable analysis and what is the difference between them? I have tried them and they seem to give different results.
  1. stcrreg i.invasion, compete(failure==2)
  2. stcrreg invasion_sup, compete(failure==2) stcrreg invasion_middle, compete(failure==2) stcrreg invasion_deep, compete(failure==2)
  3. stcrreg invasion_sup invasion_middle invasion_deep, compete(failure==2)
If I want to use stcox instead, does this change which of the above approaches I should be using or would it be the same as for stcrreg?

Many thanks for your help. Really appreciate it.