genWilcox.test: Test for Difference in Censored Samples

In USGS-R/smwrQW: Tools for censored data analysis

Test for Difference in Censored Samples

Description

This function tests for differences in censored samples from 2 groups. Two methods are available—the Peto-Prentice test is appropriate only for left-censored data. The Gehan test has been extended to multiply censored data as suggested in Gehan (1965) but uses a permutation test to compute the variance.

Usage

genWilcox.test(x, y, alternative = "two.sided", method = "best", data.names,
  gehan.seed = 0)

Arguments

x	the samples from each group. Forced to the appropriate class. Missing values are removed before the analysis.
y	either another set of samples or a group identifier with exactly two groups. Missing values are removed before the analysis. See Details.
alternative	character string describing the alternative hypothesis. Must be one of "two.sided, "greater," or "less."
method	a character string indicating which method to use for the analysis. See Details.
data.names	character string to be used to explain the data. Default names are derived from the data arguments.
gehan.seed	an integer value to set the seed to compute the variance of the Gehan statistic. If 0 (the default), then use equation 9.5 from Helsel (2012) to compute the variance.

Details

If y is either type character or factor, then it is assumed to be a group identifier. Anything else is treated as another set of sample and forced to the appropriate class of censored data.

The argument method must be one of "peto" or "gehan." It may also be "best," which selects "peto" for uncensored or left-censored data and "gehan" otherwise.

Value

An object of class "htest" that inherits "genWilcox."

Null Hypothesis

The null hypothesis is that the distributions are not different from one another.

Note

The genWilcox.test function uses the survfit function. Helsel (2012) describes flipping the left-censored data so that small values become large and left-censored values become right-censored values and adapt nonparametric techniques from survival analysis. The results from survfit are printed as the sample estimates when printing the output, the important columns are records, the number in each group; events, the number of uncensored values; and median, the group median.
A plot method is supported for the returned object.

References

Gehan, E.A., 1965, A generalized Wilcoxon test for comparing arbitraritly singly censored samples: Biometrika, v. 52, p. 203-223.

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

Helsel, D.R. 2012, Statistics for Censored Environmental Data Using Minitab and R: New York, Wiley, 324 p.

Peto, R., and Peto, J., 1972, Asymptotically efficient rank invariant test procedures (with discussion): Journal of the Royal Statistical Society, Series A v. 135, p. 185-206.

Prentice, R.L. 1978, Linear rank tests with right-censored data: Biometika, v 65, p 167-179.

Prentice, R.L., and Marke, P., 1979, A qualitative discrepancy between censored data rank tests: Biometrika, v. 35, p. 861-867.

See Also

survdiff, survfit, lcens-class

Examples

# Compare uncensored results
# First for grouped data
set.seed(69)
Xu <- rlnorm(22, 0, 1)
Yu <- rlnorm(22, .6, 1)
genWilcox.test(Xu, Yu)
wilcox.test(Xu, Yu)
# Compare effect of censoring
genWilcox.test(as.lcens(Xu, 1), Yu)
genWilcox.test(as.lcens(Xu, 1), as.lcens(Yu, 1))

USGS-R/smwrQW documentation built on Oct. 11, 2022, 6:13 a.m.