Sample rows based on weights or probability from a data.frame

Question

I would like to sample from 3 groups of data (rows) in a dataframe according to different probabilities. So for instance the data looks like

group<-c(1,1,1,2,2,3,3)
var1<-c('aa','ab','ac','ba','bb','ca','ce')
var2<-c('aaa','aba','aca','baa','bba','caa','cba')
var3<-c('aab','abb','acb','bab','bbb','cab','ceb')
data<-data.frame(group,var1,var2,var3)

Now I would like to sample (e.g. totoal of 5) from the whole dataframe (get rows) based on different probabilities (for group==1 --> 0.5, group==2 -->0.4, group==3 -->0.1)

Does anyone know how to solve that with R? Help would be much appreciated. Cheers

...

Answer 1

There is probably a more direct way via by or split and there are probably better variable names than the ones I am using, but this should work:

set.seed(4)
probs <- c(0.5, 0.4, 0.1)
grp_1Row_prob <- probs / table(data$group)
row_probs <- rep(grp_1Row_prob, times = table(data$group))
row_probs
# [1] 0.1666667 0.1666667 0.1666667 0.2000000 0.2000000 0.0500000 0.0500000

sampled_rows <- sample(1:NROW(data), size = 5, prob = row_probs)

data[sampled_rows, ]
#   group var1 var2 var3
# 2     1   ab  aba  abb
# 5     2   bb  bba  bbb
# 4     2   ba  baa  bab
# 1     1   aa  aaa  aab
# 6     3   ca  caa  cab

Answer 2

Consider using sample_n from dplyr.

library(dplyr)

group<-c(1,1,1,2,2,3,3)
var1<-c('aa','ab','ac','ba','bb','ca','ce')
var2<-c('aaa','aba','aca','baa','bba','caa','cba')
var3<-c('aab','abb','acb','bab','bbb','cab','ceb')
data<-data.frame(group,var1,var2,var3)

sample_n(data, size = 5, weight = group)

#  group var1 var2 var3
#7     3   ce  cba  ceb
#2     1   ab  aba  abb
#4     2   ba  baa  bab
#6     3   ca  caa  cab
#3     1   ac  aca  acb