根据R中的分组变量对行组重新采样

Question

I am relatively new to R so apologies if this is a silly / obvious question! I am interested in creating a new data set that is composed of collections of rows resampled with replacement from a larger data set.

The data set I have looks something like this, multiple rows per grouping variable.

> df <- data.frame(value=c(1:5,1:4,1:3),ID=c(rep(1,5),rep(2,4),rep(3,3)))
> df
   value ID
1      1  1
2      2  1
3      3  1
4      4  1
5      5  1
6      1  2
7      2  2
8      3  2
9      4  2
10     1  3
11     2  3
12     3  3

What I'd like to do is create a new data set that is resampled (with replacement) based on the grouping variable. So a resampled data set might look something like this:

   value ID
1      1  1
2      2  1
3      3  1
4      4  1
5      5  1
6      1  3
7      2  3
8      3  3
9      1  1
10     2  1
11     3  1
12     4  1
13     5  1

Thanks for any suggestions!

Answer 1

For sampling different number of rows per ID value, you can try something like this (assuming the ID value has a small number of unique values):

result <- NULL
result <- rbind(result, df[sample(row.names(df[df$ID == 1, ]), 10, replace = TRUE), ])
result <- rbind(result, df[sample(row.names(df[df$ID == 2, ]), 5, replace = TRUE), ])
result <- rbind(result, df[sample(row.names(df[df$ID == 3, ]), 3, replace = TRUE), ])
row.names(result) <- seq(1:nrow(result))

If there are many ID values, you may want to use a loop with the number of samples for each ID value you desire. For example, if there are six ID values and the corresponding numbers of samples for each ID are 10, 5, 3, 7, 8 and 2, you can do something like this:

nsamples <- c(10, 5, 3, 7, 8, 2)
result <- NULL
for (i in 1:length(nsamples)) {
  result <- rbind(result, df[sample(row.names(df[df$ID == i, ]), nsamples[i], replace = TRUE), ])
}
row.names(result) <- seq(1:nrow(result))

In either case, you will end up with output like this:

   value ID
1      1  1
2      4  1
3      1  1
4      4  1
5      2  1
6      3  1
7      1  1
8      1  1
9      4  1
10     2  1
11     2  2
12     3  2
13     1  2
14     3  2
15     1  2
16     3  3
17     2  3
18     1  3

Using the above suggested dplyr solution, you can also do something like this for variable number of samples per ID value (it also requires pre-specifying number of samples per corresponding ID in a vector):

library(dplyr)
nsamples <- c(10, 5, 3)
df %>% group_by(ID) %>% slice(sample(n(), nsamples[ID], replace = TRUE))