CODEWITHSUNDEEP
rgroup-bymedianradix
Juan
Juan

Reputation: 311

Median by group in R

I have the following data frame and would like to introduce a dummy if a value is above the group's median.

df<-data.frame(group=rep(c("A","B","c"),3), value1=c(1:9))
m<-aggregate(. ~ group, data=df, FUN=median)
names(m)[2]<-"median"
df<-merge(df,m, by="group", all.x = T)
df$median_0_1<-ifelse(df$median<df$value1,1,0)

Is there a more elegant way to do this?

And, can i adjust this to set the dummy above or below third quartile?

And, is this a robust way, that will work reliably?

Thanks a lot.

Upvotes: 1

Views: 985

Answers (2)

jay.sf
jay.sf

Reputation: 72683

Elegance lies in the eye of the beholder, but how do you like this.

df <- within(df, {
  median <- ave(value1, group, FUN=median)
  median_0_1 <- ifelse(median < value1, 1, 0)
  quantile3 <- ave(value1, group, FUN=function(x) quantile(x, probs=.3))
  quantile_0_1 <- ifelse(quantile3 < value1, 1, 0)
})
df
#   group value1 quantile_0_1 quantile3 median_0_1 median
# 1     A      1            0       2.8          0      4
# 2     B      2            0       3.8          0      5
# 3     c      3            0       4.8          0      6
# 4     A      4            1       2.8          0      4
# 5     B      5            1       3.8          0      5
# 6     c      6            1       4.8          0      6
# 7     A      7            1       2.8          1      4
# 8     B      8            1       3.8          1      5
# 9     c      9            1       4.8          1      6

Upvotes: 1

Ronak Shah
Ronak Shah

Reputation: 388907

When you want to keep number of rows in dataframe same use ave

df$median_0_1 <- with(df, as.integer(value1 > ave(value1, group, FUN = median)))

This can also be done with dplyr

library(dplyr)
df %>% group_by(group) %>% mutate(median_0_1 = as.integer(value1 > median(value1)))

Or data.table

library(data.table)
setDT(df)[, median_0_1 := as.integer(value1 > median(value1)), group]

Upvotes: 2

Related Questions