I would like to create a mtcars dataset where all cylinders have 100 observations. For that, I would sample with replacement the existing observations.
I have tried the following code that, for some reason, does not produce 300 observations.
library(data.table)
mtcars <- data.table(mtcars)
resampling <- list()
set.seed(3)
cyl <- sort(unique(as.character(mtcars$cyl)))
for (i in 1:length(cyl)){
min_obs_cyl <- 100
dat_cyl <- mtcars[cyl == as.numeric(cyl[i]) ]
resampling[[ cyl[i] ]] <- dat_cyl[sample(1:nrow(dat_cyl),
size = (min_obs_cyl - nrow(mtcars[cyl == cyl[i] ])),
replace = T),]
}
resampling_df <- do.call("rbind", resampling)
mtcars_oversample <- rbind(mtcars, resampling_df)
I get 307 observations. Anyone knows what I am doing wrong?
I think in this case, you can do the the sampling within groups using data.table's by= functionality. sample from the .I row counter within each cyl group, and then use this row identifier to sub-select the rows from the original set:
mtcars[mtcars[, sample(.I, 100, replace=TRUE), by=cyl]$V1,]
# mpg cyl disp hp drat wt qsec vs am gear carb
# 1: 18.1 6 225.0 105 2.76 3.460 20.22 1 0 3 1
# 2: 17.8 6 167.6 123 3.92 3.440 18.90 1 0 4 4
# 3: 19.2 6 167.6 123 3.92 3.440 18.30 1 0 4 4
# 4: 19.2 6 167.6 123 3.92 3.440 18.30 1 0 4 4
# 5: 21.0 6 160.0 110 3.90 2.620 16.46 0 1 4 4
# ---
#296: 15.5 8 318.0 150 2.76 3.520 16.87 0 0 3 2
#297: 19.2 8 400.0 175 3.08 3.845 17.05 0 0 3 2
#298: 19.2 8 400.0 175 3.08 3.845 17.05 0 0 3 2
#299: 14.3 8 360.0 245 3.21 3.570 15.84 0 0 3 4
#300: 15.2 8 275.8 180 3.07 3.780 18.00 0 0 3 3
If you need to specify different counts for each group, here's one option. The special .BY object stores the value of the by= argument as a list.
grpcnt <- setNames(c(50,100,70), unique(mtcars$cyl))
# 6 4 8
# 50 100 70
mtcars[mtcars[, sample(.I, grpcnt[as.character(.BY[[1]])], replace=TRUE), by=cyl]$V1]