To whom it may concern,
I read the answer to a similar question before and you guys didn’t recommend using the continuous variable Y as the response variable.
This time I have some continuous variables and want to use them as the response variables. I would like to know is there some suggested method to convert continuous variable to category variable. And how many classes (levels) for category variable the DIABLO can handle?
Looking forward to your reply!
Best regards,
Yulu
hi @yuluc ,
Apologies for the lateness in the answer. We do not usually convert continuous to categorical variables as you need to decide the cutoff optimally (and you probably saw in the literature this is not easy if you dont have expert knowledge). With DIABLO you could go with as many classes as you want, but you potentially want to focus on less than 5-8 classes to ease interpretation and number of components needed.
We have a regression version of DIABLO, called block.spls() but we are still working on the tuning parameters and implementation. If you dont need to tune the method, potentially you could use this method instead that will allow for continuous response(s).
Kim-Anh