0

I have a problem trying to fit a penalised multinomial model that helps in variable selection. The data consist of per semester record of students performances from registration to either dropout or completion. So I have multiple record per student with time being the semester registered. So a student who dropouts after 3 semesters with have the records of time as 1, 2 and 3. The data also has information about the student such as gender, whether the student perceived financial aid during the particular semester and other time varying variables and some fixed variables. Basically it is a discrete time competing survival model. However, because the simple multinomial distribution does not do variable selection, I wish to know if someone can assist me with some R code or resources that can solve this problem. I have some articles that use two penalised coefficient one for variable selection and the other for adjusting the baseline hazard but with no codes.

Thanks

lionel
  • 1
  • 1
  • Hi Lionel and welcome to stack overflow! I would update your question to be more specific to what issues you are having with R. Here are some good guidelines to help you out https://stackoverflow.com/help/how-to-ask. In general if you seeking something more in depth youtube has tons of great courses for R you can check out for free. – jmcgrath207 Aug 11 '20 at 06:25
  • I am asking what packages can be used to model penalised multinomial models for discrete time competing risks as described before. Most of the packages just for models but are not used for variable selection in those models. I am asking if someone is familiar with a package that handles penalised discrete time multinomial model that also uses a penalty term that helps in variable selection – lionel Aug 11 '20 at 21:04

0 Answers0