Add Ran requests to the presentation

[uriahf]

• How to incorporate test harm in terms of True Positives.
  Found in the original paper Decision Curve Analysis: A Novel Method for Evaluating Prediction Models.

Net benefit = TP/N - FP/N * ( pt / ( 1 - pt ) ) - test-harm

• Adjusted ROC: Under specific ( pt / ( 1 - pt ) ) draw an adjusted ROC:
  SensitivityPrevalence vs (1-Specificty)(1-Prevalence)*(pt)/(1-pt)

• Peirce, Youden, and Receiver Operating Characteristic Curves

• Basic principles of ROC analysis

• Tests that are not related to the binarization of the predictions (Like treat all)

• Relate to prevalence as not constant

• Reread all related literature (relative utility)

• Express the price of the intervention in terms of different patients (not sure how).