search

mlr-org / mlr3proba

Probabilistic Learning for mlr3
https://mlr3proba.mlr-org.com/
GNU Lesser General Public License v3.0
130 stars 20 forks source link

Suggestions: RCLL range and ERV measures properties #323

Closed bblodfon closed 1 year ago

bblodfon commented 1 year ago

Hi,

  1. Is RCLL's range correct? From it's formula, I would guess [0, Inf]?
  2. For every measure initialized with ERV = TRUE (they are all currently losses), we should also change its properties minimize (from TRUE to FALSE) and range (to [-Inf, 1] => 0 means same performance as Kaplan-Meier, negative values worse, positive values better than KM and 1 is the absolute possible maximum for ERV measures).
RaphaelS1 commented 1 year ago
  1. Good catch! Technically the upper limit actually depends on the eps parameter so with the default it's [0, 34.53878] but ignoring that yes it's [0, Inf)
  2. Also good catch!!

Can I convince you to PR these changes?

bblodfon commented 1 year ago

Can I convince you to PR these changes?

Yes, no prob, just wanted to make sure!