NRI: Net reclassification improvement

In OHDSI/PredictionComparison: Package for comparing patient level predictions developed using OHDSI's PatientLevelPreidiction package

Description

This function calculates the net reclassification improvement (NRI) metric comparing two models at a range of predicting tresholds

Usage

NRI(
  plpModel1,
  plpModel2,
  thresholds = seq(0, 1, 1/100),
  secondThresholds = NULL
)

Arguments

plpModel1	The object returned by runPlp() containing the trained model or the output when implementing an existing model
plpModel2	The object returned by runPlp() containing the trained model or the output when implementing an existing model
thresholds	A sequence of predicting tresholds to calcuate the NRI at each of these tresholds
secondThresholds	A sequence of predicting tresholds to calcuate the NRI at each of these tresholds for the second model (if NULL uses thresholds)

Details

Users need to input a trained model (the output of runPlp()) or the output of running an existing model

Value

A list containing the NRI value and z-value (can be used for statistical significance) plus various values used to calculate the NRI for each thresold. A positive value suggests the first model is better than the second model. A negative values suggests the opposite.

OHDSI/PredictionComparison documentation built on July 14, 2020, 6:16 p.m.