CODEWITHSUNDEEP
rexcelsortingmax
user001
user001

Reputation: 185

Identify and label the largest number in each group

Hi I want to identify and label the largest number for each group, can someone tell me how to get this done in r (or maybe excel would be easier)? The following is an example data, the original data contains only the left 2 columns and I want to generate the third one. In the 3rd column, I want to label the largest value in the group as 1, e.g., in group 1, the largest is .02874 so it's marked as 1, otherwise 0. Thank you!

    x <- read.table(header=T, text="group value largest
    1 0.02827 0
    1 0.02703 0
    1 0.02874 1
    2 0.03255 0
    2 0.10394 1
    2 0.03417 0
    3 0.13858 0
    3 0.16084 0
    3 0.99830 1
    3 0.24563 0")

UPDATE: Thank you all for your help! They all are great solutions!

Upvotes: 0

Views: 690

Answers (4)

Tyler Rinker
Tyler Rinker

Reputation: 109844

Here's a less cool base approach:

FUN <- function(x) {y <- rep(0, length(x)); y[which.max(x)] <- 1; y}
x$largest <- unlist(tapply(x$value, x$group, FUN))

##    group   value largest
## 1      1 0.02827       0
## 2      1 0.02703       0
## 3      1 0.02874       1
## 4      2 0.03255       0
## 5      2 0.10394       1
## 6      2 0.03417       0
## 7      3 0.13858       0
## 8      3 0.16084       0
## 9      3 0.99830       1
## 10     3 0.24563       0

It was more difficult to do in base than I had anticipated.

Upvotes: 1

flodel
flodel

Reputation: 89057

Finally, the base (no package required) approach:

is.largest <- function(x) as.integer(seq_along(x) == which.max(x))
x <- transform(x, largest = ave(value, group, FUN = is.largest))

Note that if I were you, I would remove the as.integer and just store a logical (TRUE/FALSE) vector.

Upvotes: 4

juba
juba

Reputation: 49033

Here is a solution with plyr :

x$largest <- 0
x <- ddply(x, .(group), function(df) {
    df$largest[which.max(df$value)] <- 1
    df
})

And one with base R :

x$largest <- 0
l <- split(x, x$group)
l <- lapply(l, function(df) {
    df$largest[which.max(df$value)] <- 1
    df
})
x <- do.call(rbind, l)

Upvotes: 1

TheComeOnMan
TheComeOnMan

Reputation: 12875

library(data.table)
x <- data.table(x)

y <- x[,list(value = max(value), maxindicator = TRUE), by = c('group')]

z <- merge(x,y, by = c('group','value'), all = TRUE)

Output

> z
    group   value largest maxindicator
 1:     1 0.02703       0           NA
 2:     1 0.02827       0           NA
 3:     1 0.02874       1         TRUE
 4:     2 0.03255       0           NA
 5:     2 0.03417       0           NA
 6:     2 0.10394       1         TRUE
 7:     3 0.13858       0           NA
 8:     3 0.16084       0           NA
 9:     3 0.24563       0           NA
10:     3 0.99830       1         TRUE

Upvotes: 2

Related Questions