This thread offers a solution for random multisampling with predefined probabilities. Here's an adaptation:
set.seed(1)
table(samp <- sample(1:3, size = nrow(mtcars), prob = c(0.3, 0.3, 0.4), replace = TRUE))
#>
#> 1 2 3
#> 9 8 15
nrow(mtcars) # 32
#> [1] 32
set1 = mtcars[samp == 1, ]
set2 = mtcars[samp == 2, ]
set3 = mtcars[samp == 3, ]
# disp means
mean(set1$disp)
#> [1] 263.9889
mean(set2$disp)
#> [1] 232.1625
mean(set3$disp)
#> [1] 209.9933
Created on 2022-04-29 by the reprex package (v2.0.1)
I need to take samples but in a way that disp average in each set are as close as possible.
