crrfit: 'crr' model fit statistics

In raredd/cmprsk2: cmprsk2

crr model fit statistics

Description

Functions to assess the quality of fitted crr and crr2 objects.

Select multiple types of Akaike or Bayesian (Schwarz) information criterion from a crr object to assess the relative quality of models for a given data set.

Usage

extractIC(object, ic = c("AIC", "BIC", "AICc", "BICc"), p)

## S3 method for class 'crr'
extractAIC(fit, scale, k = 2, ...)

## S3 method for class 'crr'
AIC(object, ...)

## S3 method for class 'crr'
BIC(object, ...)

## S3 method for class 'crr'
logLik(object, ...)

## S3 method for class 'crr'
deviance(object, ...)

Arguments

object	an object of class crr
ic	information criterion, one of "AIC", "BIC", "AICc", or "BICc"; see details
p	an optional penalty term to be multiplied by, k, the number of free parameters estimated in each model including the intercept term
fit, scale, k	see extractAIC
...	additional arguments passed to or from other methods

Details

AIC and BIC are calculated in the usual way. AICc is the AIC with a correction for finite sample sizes. This assumes that the model is univariate, linear, and has normally-distributed residuals (conditional upon regressors).

BICc is a newly proposed criteria that is a modified BIC for competing risks data subject to right censoring (Kuk, 2013).

References

Kuk D, Varadhan R. Model selection in competing risks regression. Stat Med. 2013 Aug 15;32.

See Also

crrFits; crrwald.test

Examples

crrs <- crr2(Surv(futime, event(censored) == death) ~ age, transplant)
crr1 <- crrs[[1L]]

crr1$coef
coef(crr1)
coefficients(crr1)

extractAIC(crr1)
sapply(crrs, AIC)

## these are equivalent
extractIC(crrs[[1L]], p = 0)
-2 * logLik(crrs[[1]])[1]
-2 * crrs[[1]]$loglik

deviance(crr1)

crrFits(crr1)
crrwald.test(crr1)

raredd/cmprsk2 documentation built on March 29, 2024, 5:34 a.m.