CODEWITHSUNDEEP
rdataframerank
Ross R
Ross R

Reputation: 389

R: Increment Rank when the column group changes

Newbie to R, I've tried Googling but I'm failing find a solution.

Here's my data frame:

Name          Value
Bob           50
Mary          55
John          51
Todd          50
Linda         56
Tom           55

So I've sorted it but what I need to add a rank column, so it looks like this:

Name          Value    Rank
Bob           50       1
Todd          50       1
John          51       2
Mary          55       3
Tom           55       3
Linda         56       4

So what I found is: 

resultset$Rank <- ave(resultset$Name, resultset$Value, FUN = rank)

But this gives me:

Name          Value    Rank
Bob           50       1
Todd          50       2
John          51       1
Mary          55       1
Tom           55       2
Linda         56       1

So close but yet so far...

Upvotes: 1

Views: 1155

Answers (3)

Frank
Frank

Reputation: 66819

Here's a base-R solution:

uv <- unique(df$Value)
merge(df,data.frame(uv,r=rank(uv)),by.x="Value",by.y="uv")

which gives

  Value  Name r
1    50   Bob 1
2    50  Todd 1
3    51  John 2
4    55  Mary 3
5    55   Tom 3
6    56 Linda 4

This is memory inefficient and has the side-effect of resorting your data. You could alternately do:

require(data.table)
DT <- data.table(df)
DT[order(Value),r:=.GRP,by=Value]

which gives

    Name Value r
1:   Bob    50 1
2:  Mary    55 3
3:  John    51 2
4:  Todd    50 1
5: Linda    56 4
6:   Tom    55 3

Upvotes: 2

Jaehyeon Kim
Jaehyeon Kim

Reputation: 1417

I guess your rank variable can be obtained by 1:length(unique(df$value)). Below is my trial.

df <- data.frame(name = c("Bob", "Mary", "John", "Todd", "Linda", "Tom"),
                 value = c(50, 55, 51, 50, 56, 55))
# rank by lengths of unique values
rank <- data.frame(rank = 1:length(unique(df$value)), value = sort(unique(df$value)))
merge(df, rank, by="value")
value  name rank
1    50   Bob    1
2    50  Todd    1
3    51  John    2
4    55  Mary    3
5    55   Tom    3
6    56 Linda    4

Upvotes: 0

A5C1D2H2I1M1N2O1R2T1
A5C1D2H2I1M1N2O1R2T1

Reputation: 193527

No need to sort... You can use dense_rank from "dplyr":

> library(dplyr)
> mydf %>% mutate(rank = dense_rank(Value))
   Name Value rank
1   Bob    50    1
2  Mary    55    3
3  John    51    2
4  Todd    50    1
5 Linda    56    4
6   Tom    55    3

Upvotes: 1

Related Questions