You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In principle, I like the idea. However, the downside is that a lot of code breaks or might break. Hence, I am not sure if we really should do this. Consider someone relying on:
attr(coef(m1), "offset")
[1] 2.922378## but with new default, we would get
attr(coef(m1, off2int=TRUE), "offset")
NULL
Similar problems might occur if someone wants to add the offset manually.
In essence, I like the idea but would still hesitate to implement it for reasons of downward compatibility.
If you really want to go that way, you should fork mboost, implement the change, make sure that the manuals and all code within mboost (also in the tests) uses the correct version of coef (i.e., with / without off2int) and finally run reverse checks to see if any other package breaks.
Shouldn't we perhaps change the default of
off2int
toTRUE
incoef.glmboost()
?The text was updated successfully, but these errors were encountered: