mets (version 1.0)

bicomprisk: Estimation of concordance in bivariate competing risks data

Description

Estimation of concordance in bivariate competing risks data

Usage

bicomprisk(formula, data, cause = c(1, 1), cens = 0, causes, indiv,
  strata = NULL, id, num, max.clust = 1000, marg = NULL,
  se.clusters = NULL, prodlim = FALSE, messages = TRUE, model,
  return.data = 0, uniform = 0, conservative = 1, resample.iid = 1, ...)

Arguments

formula

Formula with left-hand-side being a Event object (see example below) and the left-hand-side specying the covariate structure

data

Data frame

cause

Causes (default (1,1)) for which to estimate the bivariate cumulative incidence

cens

The censoring code

causes

causes

indiv

indiv

strata

Strata

id

Clustering variable

num

num

max.clust

max number of clusters in comp.risk call for iid decompostion, max.clust=NULL uses all clusters otherwise rougher grouping.

marg

marginal cumulative incidence to make stanard errors for same clusters for subsequent use in casewise.test()

se.clusters

to specify clusters for standard errors. Either a vector of cluster indices or a column name in data. Defaults to the id variable.

prodlim

prodlim to use prodlim estimator (Aalen-Johansen) rather than IPCW weighted estimator based on comp.risk function.These are equivalent in the case of no covariates.

messages

Control amount of output

model

Type of competing risk model (default is Fine-Gray model "fg", see comp.risk).

return.data

Should data be returned (skipping modeling)

uniform

to compute uniform standard errors for concordance estimates based on resampling.

conservative

for conservative standard errors, recommended for larger data-sets.

resample.iid

to return iid residual processes for further computations such as tests.

...

Additional arguments to lower level functions

Examples

data(prt) ## Prostate data example (sim)
## Bivariate competing risk, concordance estimates
p33 <- bicomprisk(Event(time,status)~strata(zyg)+id(id),
                  data=prt,cause=c(2,2),return.data=1)

p33mz <- p33$model$"MZ"$comp.risk
## Concordance
plot(p33mz,ylim=c(0,0.1),axes=FALSE); axis(2); axis(1)