RVAideMemoire (version 0.9-27)

surv.multcomp: Pairwise comparisons for survival analysis

Description

Perform pairwise comparisons between groups for equality of time to event. The function deals with coxph and survreg models.

Usage

surv.multcomp(formula, mat, data, strata = NULL, type = c("coxph", "survreg"), distribution = "exponential", p.method = "fdr")

Arguments

formula

a formula of the form a ~ b, where a is a Surv() object and b is a factor giving levels to be compared.

mat

matrix of comparisons (see help of the mat.cont() function in this package, and the esticon() function in the package doBy).

data

an optional data frame containing the variables in the formula formula. By default the variables are taken from environment(formula).

strata

strata, if needed.

type

model type: "coxph" (default) or "survreg".

distribution

distribution when a "survreg" model is used.

p.method

method for p-values correction. See help of the p.adjust() function.

Value

• model
  model formula.
• statistic
  test statistics.
• method
  a character string giving the name of the test computed.
• p.adjust.method
  method for p-values correction.
• p.value2
  corrected p-values.
• p.value
  table of results.

Details

When pairwise comparisons are based on a "coxph" model, a likelihood ratio test is used. When pairwise comparisons are based on a "survreg" model, a logrank test is used.

See Also

survreg, coxph

Examples

require(survival)
death <- c(sample(8:15,20,replace=TRUE),sample(12:20,20,replace=TRUE),sample(18:22,20,replace=TRUE))
groups <- gl(3,20,labels=LETTERS[1:3])
model1 <- coxph(Surv(death)~groups)
model1
mat <- matrix(c(1,-1,0,0,1,-1,2,-1,-1),ncol=3,byrow=TRUE,dimnames=list(1:3,levels(groups)))
mat
surv.multcomp(Surv(death)~groups,mat,type="coxph")

model2 <- survreg(Surv(death)~groups,dist="weibull")
model2
surv.multcomp(Surv(death)~groups,mat,type="survreg",distribution="weibull")