Clever Geek Handbook

R means coefficient

I have data like this

data
name v1  v2  v3  v4  v5
a    1   2   7   9   3
b    3   8   6   4   8
c    2   5   0   1   9
a    6   0   6   2   1
c    3   9   4   7   5

name

is a factor variable. I want to calculate the average by a v2,v3,v4,v5

multiplier data$name

. I used the following command, but it didn't work.

tapply(data[,3:6],data$name,mean)

I have now used the following code

newdata<-0
for (name in unique(data$name)){
    rowIndex <- which(data$name == name)
    result <- colMeans(data[rowIndex,])
    newdata[name,]<-result
}

The required result is obtained. But I want to know if there is any neat way to do this.

+3


source to share


4 answers


Here's another way

library(data.table)
cols <- paste0("v", 2:5) # set the columns you want to operate on
setDT(data)[, Sums := rowSums(.SD), .SDcols = cols]
data[, list(Means = sum(Sums)/(.N*length(cols))), by = name]
##    name Means
## 1:    a  3.75
## 2:    b  6.50
## 3:    c  5.00

Edit

Per @Aruns suggestion would probably be much better



setDT(data)[, mean(c(v2,v3,v4,v5)), by=name]
##    name   V1
## 1:    a 3.75
## 2:    b 6.50
## 3:    c 5.00

Or per @Anandas suggestion

library(reshape2)
melt(setDT(data), id.vars = "name", measure.vars = cols)[, mean(value), by = name]
##    name   V1
## 1:    a 3.75
## 2:    b 6.50
## 3:    c 5.00
+6


source


As per expected result:

t. The expected result for factor a is a (2+7+9+3)+(0+6+2+1)/8 

sapply(split(dat[,-(1:2)], dat$name), function(x) sum(x)/prod(dim(x)))
#  a    b    c 
# 3.75 6.50 5.00

or

tapply(rowMeans(dat[,-(1:2)]), dat[,1], sum)/table(dat[,1])
 #a    b    c 
 #3.75 6.50 5.00


or

 m1 <- as.matrix(dat[,-c(1:2)])
 c(by(c(m1), dat[,1][row(m1)], FUN=mean))
 #  a    b    c 
 #3.75 6.50 5.00

Or the methods suggested by @Ananda Mahto

  tapply(unlist(dat[-c(1, 2)]), rep(dat$name, 4), mean)
  #   a    b    c 
  #3.75 6.50 5.00 

  tapply(stack(dat, select = paste0("v", 2:5))$values, rep(dat$name, 4), mean)
  #  a    b    c 
  #3.75 6.50 5.00
+4


source


This can be done with a combination of the dplyr and tidyr packages:

library(dplyr)
library(tidyr)

data %>% gather(name, value, v2:v5) %>%
    group_by(name) %>% summarize(average=mean(value))
#   name average
# 1    a    3.75
# 2    b    6.50
# 3    c    5.00

This works because it gather

combines the columns v2:v5

into a single column, where they can be intuitively grouped:

data %>% gather(name, value, v2:v5)
#    name v1 name value
# 1     a  1   v2     2
# 2     b  3   v2     8
# 3     c  2   v2     5
# 4     a  6   v2     0
# 5     c  3   v2     9
# 6     a  1   v3     7
# ...
+3


source


Edit: The original answer didn't give the correct result. This seems to work fine (select (-variable) avoids having an extra column, but not required otherwise)

Using the dplyr and reshape2 packages:

library(reshape2)
library(dplyr)
data %>% 
 select(-v1) %>%
 melt %>% 
 group_by(name) %>% 
 select(-variable) %>% 
 summarise_each(funs(mean))
# Source: local data frame [3 x 2]
# 
#   name value
# 1    a  3.75
# 2    b  6.50
# 3    c  5.00
+2


source






More articles:


All Articles