wtdlogrank {PwrGSD}R Documentation

Weighted log-rank test

Description

Computes a two sample weighted log-rank statistic with events weighted according to one of the available weighting function choices

Usage

  wtdlogrank(formula = formula(data), data = parent.frame(), WtFun = c("FH", "SFH", "Ramp"),
  param = c(0, 0), sided = c(2, 1), subset, na.action, w = FALSE)

Arguments

formula a formula of the form Surv(Time, Event) ~ arm where arm is a dichotomous variable with values 0 and 1.
data a dataframe
WtFun a selection from the available list: “FH” (Fleming-Harrington), “SFH” (stopped Fleming-Harrington) or “Ramp”. See param in the following line.
param Weight function parameters. Length and interpretation depends upon the selected value of WtFun: If WtFun==\dQuote{FH} then param is a length 2 vector specifying the power of the pooled (across arms) kaplan meier estimate and its complement. If WtFun==\dQuote{SFH} then param is a length 3 vector with first two components as in the “FH” case, and third component the time (in the same units as the time to event) at which the “FH” weight function is capped off at its current value. If WtFun==\dQuote{SFH} then param is of length 1 specifying the time (same units as time to event) at which events begin to get equal weight. The “Ramp” weight function is a linearly increasing deterministic weight function which is capped off at 1 at the user specified time.
sided One or Two sided test? Set to 1 or 2
subset Analysis can be applied to a subset of the dataframe based upon a logical expression in its variables
na.action Method for handling NA values in the covariate, arm
w currently no effect

Value

An object of class survtest containing components

pn sample size
wttyp internal representation of the WtFun argument
par internal representation of the param argument
time unique times of events accross all arms
nrisk number at risk accross all arms at each event time
nrisk1 Number at risk in the experimental arm at each event time
nevent Number of events accross all arms at each event time
nevent1 Number of events in the experimental arm at each event time
wt Values of the weight function at each event time
pntimes Number of event times
stat The un-normalized weighted log-rank statistic, i.e. the summed weighted observed minus expected differences at each event time
var Variance estimate for the above
UQt Cumulative sum of increments in the sum resulting in stat above
varQt Cumulative sum of increments in the sum resulting in var above
var1t Cumulative sum of increments in the sum resulting in the variance of an unweighted version of the statistic
pu0 person units of follow-up time in the control arm
pu1 person units of follow-up time in the intervention arm
n0 events in the control arm
n1 events in the intervention arm
n sample size, same as pn
call the call that created the object

Author(s)

Grant Izmirlian <izmirlian@nih.gov>

References

Harrington, D. P. and Fleming, T. R. (1982). A class of rank test procedures for censored survival data. Biometrika 69, 553-566.

See Also

IntSurvDiff

Examples

  library(PwrGSD)
  data(lung)
  fit.wlr <- wtdlogrank(Surv(time, I(status==2))~I(sex==2), data=lung, WtFun="SFH", param=c(0,1,300))

[Package PwrGSD version 1.15 Index]