Missing metrics if maximize/minimize_boot_metric about cutpointr

[jgarces02]

Nothing, sorry for disturbing you. The problem was the number of events...

> table(u$class)
 0  1 
47  3 

[Thie1e]

Hi @jgarces02,

no worries, it's not a bad question. The error is, as you've concluded correctly, caused by the low number of positives in combination with bootstrapping, because some bootstrap sets didn't contain any positives.

You've also correctly set boot_stratify = TRUE to avoid that, but that parameter only modifies the "outer" bootstrap, not the bootstrap data sets that are used for cutpoint calculation by maximize_boot_metric. Honestly, the problem here is that we did not also implement a stratification method for the "inner" bootstrap of maximize_boot_metric or minimize_boot_metric.

If the "inner" bootstrap was also stratified, this task should run without any problems. It's of course an edge case with that high class imbalance in combination with the low sample size, but I think we should be able to handle that. I'm going to add this to my to do list.

Best,
Christian