wkm.Stat: Weighted Kaplan-Meier Test Statistic
In keaven/nphsim: Non proportional hazards sample size and simulation

Description Usage Arguments Details Value Examples

Produce the Weighted Kaplan-Meier test statistic. Test statistic is constructed based on the paper of Pepe and Fleming (1989). Coding is based on work from a non-proportional hazards working group at Merck; original author ??

1	wkm.Stat(survival, cnsr, trt, stra = NULL, fparam = NULL)

`survival`	time-to-event variable
`cnsr`	censoring variable: 1=censoring, 0=event
`trt`	treatment varaible. Accepted values are either "experiment" or "control"
`stra`	stratification variable. Default is `NULL` (currently not implemented)
`fparam`	parameter description. Set to `NULL` (currently not used)

Although it is of interest that the statistic will reduce to RMST when the weight function = 1, the stability conditions proposed by Pepe and Fleming (1989) rule out a constant weight function when censoring is present. Input: dataframe 'indata' contains the following variables: Event.T -> Event time (failure/censor); Status -> 0 = Censored, 1 = Event of Interest; Z -> 0 = Control, 1 = Active

Code returns the Z statistic and p-value.

pval: One-sided p-Value from weighted Kaplan-Meier test
z: test statistics
Df: degree of freedom. Currently always set to NA

# weighted Kaplan-Meier test on the simulated data
library(survival)
medC = 6 
hr <- c(1, 0.6)
intervals <- 3 
gamma <- c(2.5, 5,  7.5,  10) ## a ramp-up enrollment
R     <- c(2  , 2,  2  ,  6 ) ## enrollment period: total of 12 months
eta <- -log(0.99) ## 1% monthly dropout rate
sim1 <- nphsim(nsim=1,lambdaC=log(2)/medC,lambdaE=log(2)/medC*hr, ssC=300,ssE=300,
               intervals=intervals,gamma=gamma, R=R,eta=eta)
test1 <- simtest(x=sim1, anaD=c(250,300), method=wkm.Stat)
test1$result[]

# direct function call (without cutoff)
wkm.Stat(surv=sim1$simd$survival, cnsr=sim1$simd$cnsr, trt=sim1$simd$treatment)