I need to report difference in median (or Median difference) and its 95% confidence interval-CodePudding

I have 2 unequal groups (pp and Control) and I reported median and 95% confidence interval for waiting.time variable for each group but now I want to report "Difference in median" and its 95% confidence interval. I found something similar but in sas (

Any advice will be greatly appreciated

CodePudding user response：

To bootstrap the difference in medians and then compute quantiles can be done as follows.

1.

First read in the data.

data <- read.csv("~/Transferências/stackoverflow.data.csv")
str(data)
#'data.frame':  420 obs. of  3 variables:
# $ X           : int  1 2 3 4 5 6 7 8 9 10 ...
# $ group       : chr  "Control" "Control" "PP" "Control" ...
# $ waiting.time: num  NA NA NA 23.9 NA ...

The data has NA's and argument na.rm = TRUE will be used.

2.

Now bootstrap the statistic, function boot_diff_median.

library(boot)

boot_diff_median <- function(data, i){
  diff(tapply(data$waiting.time[i], data$group[i], FUN = median, na.rm = TRUE))
}

set.seed(2021)
R <- 1e4

b <- boot(data, statistic = boot_diff_median, R = R)

3.

Compute quantiles. I have also included the bootstrapped value, mean(b$t).

mean(b$t)
#[1] -0.4330915

t(quantile(b$t, probs = c(0.025, 0.5, 0.975), na.rm = TRUE))
#          2.5%  50%    97.5%
#[1,] -11.35525 -0.5 8.620875