Closed RaphaelS1 closed 9 months ago
What's the status here? I'd like to try the difference method soonish -- should I rather test-drive the PR first?
Just notice that our issue of getting Inf
here doesn't go away because the issue appears to be in
out = deaths / sum(haz)
where for some reason sum(haz)
appears to be 0, so
out = abs(1 - out)
will still give us an Inf
.
I may also have forgotten what we get by using the difference method but I had just been staring at the Inf
s in my bma
and noticed this 😅
logloss
optimized to use survival matricesbeta_calib
diff version testedalpha_calib
optimized and diff version testedInf
in calib_alpha
=> fix: use eps
Inf
in graf improper => bug fix: remove Inf
's from integration methodsaddresses #350 and #365
Added methods so calib alpha and beta can be minimised