survival task chapter

[adibender]

• make clearer distinction between distribution predictions Distr(Y) and scalar prediction E(Y)
• relationship between relative risk, tte, prognostic index. Which is the most general? Can the others be viewed as special cases?
• Ranking example -> data should not be in correct order already
• in "Predicting survival time": reconsider if speaking of losses is necessary. Potentially simplify ("It might just be easier to say that because of censoring we often don't even know Y in the training data so even training is probabilistic in nature. Or if you do want to talk about losses then more simply 'evaluating predicting time to events is difficult as test data still includes censoring'")