Subset a dataset based on the appearance of values in a column in R-CodePudding

I have a dataset (ds) with two columns. There are either one or two entries with the same number in "match". "status" is a binary variable. There are pairs, for example, the value 12 in match appears twice, one for where status is 1 and 0 for the other. Yet, there are also observations in match who do not have a partner, in this dataset it would be 3,8,33,17 who have no partner.

match     status
12          1 
3           1
5           0
8           1
33          0
5           1
12          0
17          0

What I want to do is to create a new dataset that only contains observations of pairs (thus if a value appears twice). In my example, it would be

match     status
12          1
12          0
5           0
5           1

The status variable in the final dataset would be 50/50 because a value in match (for example 12) has an observation where status = 0 and one where status = 1. The actual dataset I´m working with has over 50k observations so I cannot just search and filter by each number. What I tried is:

numbers <- table(ds$match)
numbers <- as.data.frame(numbers)
numbers <- numbers[numbers$Freq == 2,]
numbers <- numbers$Var1
ds$keep <- ifelse(numbers %in% ds$match, 1, 0)

Here I get the error "replacement has 23005 rows, data has 39021" If I could get around this error I think I could just run

ds <- filter(ds, ds$keep == 1)

to get the dataset that I want. This was my most promising approach. I tried a few other things but it always came done to the fact that the status variable wasn´t 50/50 so I couldn´t manage to exclude all observations without a pair. Does someone have an idea how I could fix my code or is there a solution that would be quicker/more smooth? Thanks for any help in advance!

CodePudding user response：

library(dplyr)

ds %>% group_by(match) %>% filter(n()>1) %>% arrange(match,status)

  match status
  <dbl>  <dbl>
1     5      0
2     5      1
3    12      0
4    12      1

You can also do something like this:

ds <- ds[order(ds$match),]
id = rle(ds$match)
ds[ds$match %in% id$values[id$lengths>1],]

  match status
  <dbl>  <dbl>
1     5      0
2     5      1
3    12      1
4    12      0