reclass: Retrospective risk reclassification table

In pec: Prediction Error Curves for Risk Prediction Models in Survival Analysis

Retrospective risk reclassification table

Description

Retrospective table of risks predicted by two different methods, models, algorithms

Usage

reclass(
  object,
  reference,
  formula,
  data,
  time,
  cause,
  cuts = seq(0, 100, 25),
  digits = 2
)

Arguments

object	Either a list with two elements. Each element should either be a vector with probabilities, or an object for which predictSurvProb or predictEventProb can extract predicted risk based on data.
reference	Reference prediction model.
formula	A survival formula as obtained either with prodlim::Hist or survival::Surv which defines the response in the data.
data	Used to extract the response from the data and passed on to predictEventProb to extract predicted event probabilities.
time	Time interest for prediction.
cause	For competing risk models the cause of interest. Defaults to all available causes.
cuts	Risk quantiles to group risks.
digits	Number of digits to show for the predicted risks

Details

All risks are multiplied by 100 before

Value

reclassification tables: overall table and one conditional table for each cause and for subjects event free at time interest.

Author(s)

Thomas A. Gerds <[e-mail]>

See Also

predictStatusProb

Examples

## Not run: 
library(survival)
set.seed(40)
d <- prodlim::SimSurv(400)
nd <- prodlim::SimSurv(400)
Models <- list("Cox.X2"=coxph(Surv(time,status)~X2,data=d,x=TRUE,y=TRUE),
               "Cox.X1.X2"=coxph(Surv(time,status)~X1+X2,data=d,x=TRUE,y=TRUE))
rc <- reclass(Models,formula=Surv(time,status)~1,data=nd,time=5)
print(rc)
plot(rc)

set.seed(40)
library(riskRegression)
library(prodlim)
dcr <- prodlim::SimCompRisk(400)
ndcr <- prodlim::SimCompRisk(400)
crPred5 <- list("X2"=predictEventProb(CSC(Hist(time,event)~X2,data=dcr),newdata=ndcr,times=5),
                "X1+X2"=predictEventProb(CSC(Hist(time,event)~X1+X2,data=dcr),newdata=ndcr,times=5))
rc <- reclass(crPred5,Hist(time,event)~1,data=ndcr,time=3)
print(rc)

reclass(crPred5,Hist(time,event)~1,data=ndcr,time=5,cuts=100*c(0,0.05,0.1,0.2,1))

## End(Not run)

pec documentation built on April 11, 2023, 5:55 p.m.