AHR (version 1.5.0)

ahrUser: ahrUser

Description

Estimate average hazard ratios from k independent samples based on user-supplied survival function estimator

Usage

ahrUser(L, formula, data, null.theta = NULL, contrast = NULL,
  multi.test = FALSE, cov = FALSE, bootstrap = 0,
  user.survfit = wkm, user.param = list(alpha = 1, var = FALSE, cov =
  FALSE, left.limit = FALSE, rr.subset = rep(TRUE, nrow(data))))

Arguments

L

time-limit specifying time-interval [0,L] over which average hazard ratios will be calculated

formula

an object of class '"formula"' specifying the conditional survival model

data

data frame containing the variables in formula

null.theta

vector specifying the null hypothesis for the average hazard ratios (H_0: theta = null.theta)

contrast

vector of contrasts to test H_0: contrast * (theta - null.theta) = 0

multi.test

calculate multivariate test statistic if TRUE

cov

if TRUE calculate covariance matrix estimator (direct)

bootstrap

if > 0 then use bootstrap to estimate covariance matrix (ignore if cov is TRUE)

user.survfit

user defined function taking vector of times, data.frame and list of parameters returning survival function estimate

user.param

list of parameters passed to function user.survfit

Value

An object of class '"ahr"'

Details

user.survfit must return logV (if user supplied survival function estimator hast independent increments property) or logCOV

References

J.~D. Kalbfleisch and R.~L. Prentice. Estimation of the average hazard ratio. Biometrika, 68(1):105--112, Apr. 1981.

See Also

wkm

Examples

# NOT RUN {
## User supplied survival function estimator (should be exactly the same as
T <- c(rexp(100, 1), rexp(100, 2))
C <- c(rexp(100, 1), rexp(100, 2))
time <- pmin(T, C)
status <- T <= C
trt <- rep(c(0,1), c(100, 100)) # treatment indicator

sfit <- function(times, data, param) {
  fit <- survfit(Surv(Y, D) ~ 1, data=data)
  f <- approxfun(fit$time, fit$surv, method="constant", f=0, yleft=1, rule=2)
  fv <- approxfun(fit$time, fit$std.err^2, method="constant", f=0, yleft=1, rule=2)

  S <- f(times)
  logV <- fv(times) * nrow(data)
  V <- S^2 * logV

  list(times=times, S=S, V=V, logV=logV)
}
fit1 <- ahrUser(2, Surv(time, status) ~ trt,
                data.frame(time=time, status=status, trt=trt), user.survfit=sfit, user.param=list())
fit1
fit2 <- ahrKM(2, Surv(time, status) ~ trt, data.frame(time=time, status=status, trt=trt), cov=FALSE)
fit2

# }