Rank doesn't start at 1 in R

Question

I want to create a rank variable

Setup

test <- data.frame(column1 = c(5,5,5,6,6,7,7,7,8))
test$rank <- rank(test)

 test
  column1 rank
1       5  2.0
2       5  2.0
3       5  2.0
4       6  4.5
5       6  4.5
6       7  7.0
7       7  7.0
8       7  7.0
9       8  9.0

The answer I want is 1,1,1,2,2,3,3,3,4.

The answer you want isn't a rank. If the first 3 entries are equal, then the next 2 can't be equal "2" as "2" and "3" belong to the first 3. You can play around with the `ties.method` in order to check different options. In order to reach your desired output, you could do something like `match(test$column1, sort(unique(test$column1)))` — David Arenburg, Apr 19 '20 at 07:55

nikn8 · Accepted Answer · 2020-04-19T08:15:46.260

2

You need to use dense_rank.

test <- data.frame(column1 = c(5,5,5,6,6,7,7,7,8))
test$rank <- dplyr::dense_rank(test$column1)

Working of window ranking function

test %>% rename(input = column1) %>% 
  mutate(row_num_output = row_number(input),
                rank_output = min_rank(input),
                dense_rank_output = dense_rank(input))

Output to give a better understanding for your input

edited Apr 19 '20 at 08:15

answered Apr 19 '20 at 07:55

nikn8

1,016
8
23

[Here](http://www.besttechtools.com/articles/article/sql-rank-functions), One of the best place to understand window ranking function. it's in SQL but concepts remain same, regardless of language. – nikn8 Apr 19 '20 at 08:06

rg255 · Answer 2 · 2020-04-19T08:25:54.183

0

A data.table solution, using frank() (fast rank) function which has the ties method "dense".

library(data.table)
test <- data.table(column1 = c(5,5,5,6,6,7,7,7,8))
test[, rank := frank(column1, ties.method = "dense")]

Alternatively a base R solution using match

test$rank <- match(test$column1, unique(test$column1[order(test$column1)]))

edited Apr 19 '20 at 08:25

answered Apr 19 '20 at 08:00

rg255

4,119
3
22
40

score 0 · Answer 3 · answered Apr 19 '20 at 08:14

There are multiple ways you can do this :

In dplyr, you can use group_indices

test$rank <- dplyr::group_indices(test, column1)

Or in base R, cumsum with duplicated.

test$rank <- cumsum(!duplicated(test$column1))

Make sure column1 is ordered before you use the above since both the methods are sensitive to order.

Rank doesn't start at 1 in R

3 Answers3

Linked

Related