Return multiple values from a function by column and group in R

Question

I am trying to find the proportion of data that is greater than 20 for each of the factors I have in the data frame, then use those proportions to compute 2 other values:

dat <- data.frame(num1=as.numeric(c(10,30,4,60,20,1,34,87,66)), num2=as.numeric(c(23,36,42,18,3,44,32,65,78)), num3=as.numeric(c(0,0,0,20,80,10,50,43,70)), group=c("First group", "First group","First group", "Second group","Second group","Second group", "Third group","Third group","Third group"))

I would like to get 3 values (from a function) computed for each of the columns num1, num2 and num3, and each of the groups like this:

res = data.frame(cbind(col=c(rep("num1",3), rep("num2",3), rep("num3",3)), group=rep(c("First group", "Second group","Third group"),3) , p= c(0.3333333, 0.3333333, 1.0000000,1.0000000, 0.3333333,1.0000000,0.0000000,0.3333333,1.0000000), s1= c(-0.1250000, -0.1250000, -0.2500000,-0.2500000,-0.1250000,-0.2500000,0.0000000,-0.1250000,-0.2500000), s2= c(0.1000000, 0.1000000, 0.5000000,0.5000000, 0.1000000, 0.5000000, 0.0000000,0.1000000,0.5000000)))

I can get as far as returning data for each column like this:

prop <- function(s) {
 n= length(s)
 x=length(s[s>20])
 p=x/n
 s1=(p/2-p)/(p+1)
 s2=(p/2-p)/(p-2)
 return(c(p,s1,s2))
 }

ddply(dat, .(group), summarise, prop(num1))

but then I don't understand how to bind them into a dataframe and apply to each columns. I have tried different ways (for example this but it is not working for me as I keep getting only one column. I am trying to do this by the way to then plot these values by group using ggplot2. Can you please help me?

score 1 · Accepted Answer · answered Feb 12 '14 at 13:05

prop <- function(s) {
  n= length(s)
  x=length(s[s>20])
  p=x/n
  s1=(p/2-p)/(p+1)
  s2=(p/2-p)/(p-2)
  data.frame(p,s1,s2)
}

library(reshape2)
dat <- melt(dat, id="group")
library(plyr)
ddply(dat, .(variable, group), function(df) prop(df$value))

#  variable        group         p     s1  s2
#1     num1  First group 0.3333333 -0.125 0.1
#2     num1 Second group 0.3333333 -0.125 0.1
#3     num1  Third group 1.0000000 -0.250 0.5
#4     num2  First group 1.0000000 -0.250 0.5
#5     num2 Second group 0.3333333 -0.125 0.1
#6     num2  Third group 1.0000000 -0.250 0.5
#7     num3  First group 0.0000000  0.000 0.0
#8     num3 Second group 0.3333333 -0.125 0.1
#9     num3  Third group 1.0000000 -0.250 0.5

Aaaa yes, changing to long format! Thank you very much! – user2337032 Feb 12 '14 at 13:24 — user2337032, Feb 12 '14 at 13:24

score 1 · Answer 2 · answered Feb 12 '14 at 13:21

A Solution without additional packages would be:

s1<-function(p){(p/2-p)/(p+1)}
s2<-function(p){(p/2-p)/(p-2)}

dat.split <- split(dat,dat$group)
L<-lapply(dat.split,function(data){
  group<-data[,1:3]
  p1<-sum(group$num1>20)/nrow(group)
  p2<-sum(group$num2>20)/nrow(group)
  p3<-sum(group$num2>20)/nrow(group)
  tmp<-c(p1,p2,p3)
  return(data.frame(name=c("num1","num2","num3"),
                    group=data[,4],
                    prob=tmp,
                    stat1=sapply(tmp,s1),
                    stat2=sapply(tmp,s2)))
})

do.call("rbind", L)

Thank you so much to both, answers are both exactly what I was looking for! Just checking the first answer I got, but this is perfect too. — user2337032, Feb 12 '14 at 13:24

Return multiple values from a function by column and group in R

2 Answers2

Linked