BuyseTest (version 3.0.5)

CasinoTest: Multi-group GPC (EXPERIMENTAL)

Description

Perform Generalized Pairwise Comparisons (GPC) for two or more groups. Can handle one or several binary, continuous and time-to-event endpoints.

Usage

CasinoTest(
  formula,
  data,
  type = "unweighted",
  add.halfNeutral = NULL,
  method.inference = "u-statistic",
  conf.level = NULL,
  transformation = NULL,
  alternative = NULL,
  method.multcomp = "none",
  seed = NA
)

Value

An S3 object of class CasinoTest that inherits from data.frame.

Arguments

formula

[formula] a symbolic description of the GPC model, see the BuyseTest function

data

[data.frame] dataset.

type

[character] Type of estimator: can be "unweighted" or "weighted".

add.halfNeutral

[logical] should half of the neutral score be added to the favorable and unfavorable scores?

method.inference

[character] method used to compute confidence intervals and p-values. Can be "none", "u-statistic", or "rank".

conf.level

[numeric] confidence level for the confidence intervals. Default value read from BuyseTest.options().

transformation

[logical] should the CI be computed on the inverse hyperbolic tangent scale / log scale for the net benefit / win ratio and backtransformed. Otherwise they are computed without any transformation. Default value read from BuyseTest.options(). Not relevant when using permutations or percentile bootstrap.

alternative

[character] the type of alternative hypothesis: "two.sided", "greater", or "less". Default value read from BuyseTest.options().

method.multcomp

[character] method used to adjust for multiple comparisons. Can be any element of ‘p.adjust.methods’ (e.g. "holm"), "maxT-integration", or "maxT-simulation".

seed

[integer, >0] Random number generator (RNG) state used when adjusting for multiple comparisons. If NULL no state is set.

Details

Require to have installed the package riskRegression and BuyseTest

Setting argument method.inference to "rank" uses a U-statistic approach with a small sample correction to match the variance estimator derived in Result 4.16 page 228 of Brunner (2018).

References

Edgar Brunner, Arne C Bathke, and Frank Konietschke (2018). Rank and pseudo-rank procedures for independent observations in factorial designs. Springer.

Examples

library(data.table)
library(BuyseTest)

#### simulate data ####
set.seed(11)
n <- 4
dt <- rbind(data.table(score = rnorm(n), group = "A"),
            data.table(score = rnorm(2*n), group = "B"),
            data.table(score = rnorm(3*n), group = "C"))
dt$index <- 1:NROW(dt)

#### estimation ####
score.casino <- dt$score

## naive casino (by hand)
M.score <- outer(dt[group=="A",score],score.casino,function(x,y){x>y+0.5*(x==y)})
mean(M.score)

## naive casino (via BuyseTest)
CasinoTest(group ~ cont(score), data = dt, type = "weighted")

## harmonic casino (by hand)
hweight <- unlist(tapply(dt$group, dt$group, function(x){rep(1/length(x),length(x))}))
M.scoreW <- sweep(M.score, MARGIN = 2, FUN = "*", STATS = NROW(dt)*hweight/3)
mean(M.scoreW)

## harmonic casino (via BuyseTest)
CasinoTest(group ~ cont(score), data = dt, type = "unweighted")

#### Relative liver weights data (Brunner 2018, table 4.1, page 183) ####
liverW <- rbind(
  data.frame(value = c(3.78, 3.40, 3.29, 3.14, 3.55, 3.76, 3.23, 3.31),
             group = "Placebo"),
  data.frame(value = c(3.46,3.98,3.09,3.49,3.31,3.73,3.23),
             group = "Dose 1"),
  data.frame(value = c(3.71, 3.36, 3.38, 3.64, 3.41, 3.29, 3.61, 3.87),
             group = "Dose 2"),
  data.frame(value = c(3.86,3.80,4.14,3.62,3.95,4.12,4.54),
             group = "Dose 3"),
  data.frame(value = c(4.14,4.11,3.89,4.21,4.81,3.91,4.19, 5.05),
             group = "Dose 4")
)
liverW$valueU <- liverW$value + (1:NROW(liverW))/1e6

## same as table 4.1, page 183 in Brunner et al (2018)
CasinoTest(group ~ cont(value), data = liverW, type = "weighted", add.halfNeutral = TRUE)
CasinoTest(group ~ cont(valueU), data = liverW, type = "unweighted", add.halfNeutral = TRUE)