Clever Geek Handbook

Hence, all other columns are based on the same column in r

I have a large data frame that contains over 40,000 columns and I am facing a problem like this Sum over different column value in R

shop <- data.frame( 
  'shop_id' = c('Shop A', 'Shop A', 'Shop A', 'Shop B', 'Shop C', 'Shop C'), 
  'Assets' = c(2, 15, 7, 5, 8, 3),
  'Liabilities' = c(5, 3, 8, 9, 12, 8),
  'sale' = c(12, 5, 9, 15, 10, 18), 
  'profit' = c(3, 1, 3, 6, 5, 9))

I have a shop_id column that is repeated many times. I have other values ​​associated with this shop_id, such as assets, liabilities, profit, loss, etc. Now I want to average over all variables that have the same shop_id, that is, I want unique shop_ids and I want to average all columns that have the same shop_id. Because thousands of variables (columns) working with each column (variable) separately is very tedious.

My answer should be

 shop_id  Assets  Liabilities     sale    profit    
 Shop A   8.0     5.333333    8.666667  2.333333
 Shop B   5.0     9.000000   15.000000  6.000000
 Shop C   5.5    10.000000   14.000000  7.000000

I currently use nested loops for the following: As generic as R, I believe there must be a faster way to do this

idx <- split(1:nrow(shop), shop$shop_id)

newdata <- data.frame()

for( i in 1:length(idx)){
    newdata[i,1]<-c(names(idx)[i] )
    for (j in 2:ncol(shop)){
        newdata[i,j]<-mean(shop[unlist(idx[i]),j])
    }
}
+3


source to share


4 answers


Try data.table 

library(data.table)
setDT(shop)[, lapply(.SD, mean), shop_id]
#  shop_id Assets Liabilities      sale   profit
#1:  Shop A    8.0    5.333333  8.666667 2.333333
#2:  Shop B    5.0    9.000000 15.000000 6.000000
#3:  Shop C    5.5   10.000000 14.000000 7.000000

or

library(dplyr)
shop %>% 
    group_by(shop_id)%>%
    summarise_each(funs(mean))
# shop_id Assets Liabilities      sale   profit
#1  Shop A    8.0    5.333333  8.666667 2.333333
#2  Shop B    5.0    9.000000 15.000000 6.000000
#3  Shop C    5.5   10.000000 14.000000 7.000000


or

aggregate(.~shop_id, shop, FUN=mean)
#   shop_id Assets Liabilities      sale   profit
#1  Shop A    8.0    5.333333  8.666667 2.333333
#2  Shop B    5.0    9.000000 15.000000 6.000000
#3  Shop C    5.5   10.000000 14.000000 7.000000

For 40,000 columns, I would data.table

or might be dplyr

.

+3


source


Try with dplyr

:



library("dplyr")
shop %>% group_by(shop_id) %>% summarise_each(funs(mean))

#   shop_id Assets Liabilities      sale   profit
# 1  Shop A    8.0    5.333333  8.666667 2.333333
# 2  Shop B    5.0    9.000000 15.000000 6.000000
# 3  Shop C    5.5   10.000000 14.000000 7.000000
+2


source


rowsum

might be useful here too:

rowsum(shop[-1], shop[[1]]) / table(shop[[1]])
#       Assets Liabilities      sale   profit
#Shop A    8.0    5.333333  8.666667 2.333333
#Shop B    5.0    9.000000 15.000000 6.000000
#Shop C    5.5   10.000000 14.000000 7.000000
+2


source


Use a ddply

function from the package plyr

:

> require("plyr")
> ddply(shop, ~shop_id, summarise, Assets=mean(Assets),
        Liabilities=mean(Liabilities), sale=mean(sale), profit=mean(profit))

  shop_id Assets Liabilities      sale   profit
1  Shop A    8.0    5.333333  8.666667 2.333333
2  Shop B    5.0    9.000000 15.000000 6.000000
3  Shop C    5.5   10.000000 14.000000 7.000000
+1


source






More articles:


All Articles