How to conduct a boostrapped or a permutation wilcoxon test (in R)

Question

I have a total of 18 mice, 9 from Healthy group and 9 from Sick group. Since I have a very poor sample size I would like to use boostrapping or permutation test.

Is there a way to conduct a boostrapped or a permutation wilcoxon rank sum test in R ? thus obtaining a boostrapped p-value or p value from permutation test

For now I just used a classical wilcoxon rank sum :

wilcox_test(DV~ Group,paired = FALSE,data=df)
A tibble: 1 × 7
.y.                     group1 group2    n1    n2 statistic     p

<chr>                   <chr>  <chr>  <int> <int>     <dbl> <dbl>

1  DV                     G1     G2         9     9        51 0.387

I tried this (boostrapping) :

boot_wilcoxon<- function(formula, data, indices) {
    d <- data[indices,] # allows boot to select sample
    fit <- wilcox_test(formula, data=d)
    return(fit$statistic)
}
library(boot)
results <- boot(data=df, statistic=boot_wilcoxon,
                R=1000, formula=DV~ Group)

to get the p-value:
sum(results$t>results$t0)/1000
[1] 0.406

Does it look ok? (also the boostrapped p value seems closed to the original p-value)

Bootstrap your sample, perform the test, save the statistic/p-value, repeat. — user2974951, Jun 27 '22 at 11:38
Yes @user2974951 but how to get the final boostrapped p value ? Is it the number of time where the p value is <0.05 devided by the number of repetition ? — learners, Jun 27 '22 at 11:38
The number of times the original p-values is < the simulated p-values, however... see Bootstrap vs. permutation hypothesis testing — user2974951, Jun 27 '22 at 12:00
@user2974951 Yes, for the permutation test, I am reading "A simple calculation for the p-value of the permutation test is to count the number of statistics from permuted data that exceed the statistic for the original data and then divide by the number of permutations performed." I presume it's the same for bootstrapping ? (I also edited my post) — learners, Jun 27 '22 at 12:11

How to conduct a boostrapped or a permutation wilcoxon test (in R)

A tibble: 1 × 7

I tried this (boostrapping) :

to get the p-value:

0 Answers0