This is an R package for quickly approximating small p-values (e.g. p < 10-6) for the difference and ratio of means in two-sample tests as described by Segal et al. (2018).
This package also implements stochastic approximation Monte Carlo (Yu et al., 2011) in the SAMC function.
# install.packages("devtools")
library("devtools")
install_github("bdsegal/fastPerm")
library(fastPerm)
# Ratio of means
x <- rexp(100, 5)
y <- rexp(100, 2)
# Expected stopping partition, based on asymptotic approximation
mStopRatioMean(x, y)
# Fast permutation test, defaults to 'testStat = ratioMean'
fastPerm(x, y)
# Difference of means
x <- rnorm(110, 0, 1)
y <- rnorm(95, 1, 1)
# Expected stopping partition, based on asymptotic approximation
mStopDiffMean(x, y)
# Fast permutation test
fastPerm(x, y, testStat = diffMean)
Segal, B. D., Braun, T., Elliott, M. R. and Jiang, H. (2018). Fast approximation of small p-values in permutation tests by partitioning the permutations. Biometrics, 74(1), 196-206. doi:10.1111/biom.12731.
Yu, K., Liang, F., Ciampa, J., and Chatterjee, N. (2011). Efficient p-value evaluation for resampling-based tests. Biostatistics 12, 582–593.