row_number versus min_rank

dplyr-minrank.R

# generate some test data
set.seed(100)
test <- tbl_df(data.frame(ID=sample(1:100, 100000, replace=TRUE), X=sample(1:1000, 100000, replace=TRUE)))
test <- arrange(test, ID)
people <- group_by(test, ID)

mutate(people, new=row_number(desc(X))) %.% filter(ID==1) %.% arrange(new)
# result looks good:
#ID   X new
#1   1 999   1
#2   1 997   2
#3   1 996   3
#4   1 995   4
#5   1 995   5
#6   1 995   6
#7   1 994   7
#8   1 993   8
#9   1 993   9
#10  1 993  10

mutate(people, new=min_rank(desc(X))) %.% filter(ID==1) %.% arrange(new)
# result looks good:
#ID   X new
#1   1 999   1
#2   1 997   2
#3   1 996   3
#4   1 995   4
#5   1 995   4
#6   1 995   4
#7   1 994   7
#8   1 993   8
#9   1 993   8
#10  1 993   8

mutate(people, new=row_number(desc(X))) %.% filter(ID==2) %.% arrange(new)
# result looks good:
#ID    X new
#1   2 1000   1
#2   2  997   2
#3   2  997   3
#4   2  997   4
#5   2  996   5
#6   2  996   6
#7   2  996   7
#8   2  994   8
#9   2  993   9
#10  2  993  10

mutate(people, new=min_rank(desc(X))) %.% filter(ID==2) %.% arrange(new)
# strange result:
#ID   X new
#1   2 204   1
#2   2 451   2
#3   2 413   3
#4   2 688   4
#5   2 676   4
#6   2 988   6
#7   2 867   7
#8   2 154   7
#9   2 773   9
#10  2 541  10

### For every ID I have tried with these pair of statements (other than ID==1),
### row_number() seems to work and min_rank() doesn't.