0

Well, im trying to fit a model with 5 "classes" Controls - DiseaseA_soft - DiseaseA_hard - DiseaseB_soft - DiseaseB_hard.

Being soft and hard 2 levels of their prognosis. "Hard" is way severe than "soft". Anyway, i don think a classical "Class ~." model should be used, cause we got 2 sub-ordinal levels for each disease. ¿Is there a way i can correctly represent this "design" with nnet::multinom?

For a 5 classes (clase = 1 column with 5 levels of a factor unordered) is just

multinom(clase~variables, data= data_training, maxit=10000 )

I dont know if i should make 2 columns, 1 with a 3 lvl factor determining Controls, Dis_A, and Dis_B, and the other with the severity of the disease. (Mostly because i dont know what to do with the Controls, i dont think is correct to treat them as the "gentle" version of the disease.

¿Can i do something like this? ¿And that will be correct?

multinom(disease_type+severity~variables, data= data_training, maxit=10000 )
Galpaccru
  • 57
  • 12
  • It's considered bad form to cross post to multiple stack exchange sites. Without some sort of [reproducible example](https://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example) with sample input and desired output that can be used to test and verify possible solutions, this really isn't a question that's appropriate for Stack Overflow. – MrFlick Mar 30 '20 at 19:44
  • Thank you, im new to this community. Before deleting, is my question appropiate for stats then? – Galpaccru Mar 30 '20 at 19:49
  • It is now correct and appropiate for stackoverflow @MrFlick? – Galpaccru Mar 31 '20 at 07:50

0 Answers0