Open PhilippBach opened 10 months ago
@PhilippBach I think it makes sense to require that poly >1 -- because that is the case when classic linear model is used, which would likely the most simple model that applied people would want to try I think. I will be working on the package today - trying to figure out the data.table warning
@klosins and I agreed that allowing for poly == 1
would implement a standard benchmark, so we should allow and handle this case explicitly
One thing we should take care of is if the poly_order is set to 0 ... Then this loop here doesn't make sense
https://github.com/MartinSpindler/hdm/blob/d9fe0f09a39a32959f0344632b0dda25197b9dd2/R/AutoDebiased_backends.R#L178
@klosins - do you think it makes sense to generally require
poly_order > 1
. Otherwise, the derivative is constant anyways...Otherwise, we'd have to include another
if (poly_order == 1)
condition here