根据R中另一列的条件求和

Question

I have a dataset with 150+ columns and 1000s of rows. The dataset provides a flag for different items for various categories in different columns. One of the columns is total usage for each item across the category. Below is sample of the dataset:

Values   A B C
1        Y   
2          Y
3        Y   Y 
4            Y 

I want to use R to do calculations such that I get the following results:

     Count  Sum
A      2     4
B      1     2
C      2     7

Basically I want the Count Column to give me the number of "y" for A, B and C, and the Sum column to give me sum from the Usage column for each time there is a "Y" in Columns A, B and C

Step 2 - I have similar column values in 200 + files. I have brought all the files into a folder. What I would like to do is use the above functions, apply it in each of the file, and then have the answer grouped by file and category. for example

File 1 Count A Sum A Count B Sum B Count C Sum C

File 2 Count A Sum A Count B Sum B Count C Sum C

and so on

Answer 1

Here's one simple (step-by-step) solution

# First, readind your data
> df <- read.table(text="Values   A  B  C
+ 1        Y  NA NA
+ 2        NA Y  NA
+ 3        Y  NA Y 
+ 4        NA NA Y ", header=TRUE)
> 
> Count <- colSums(!is.na(df[, -1]))
> Sum <- apply(!is.na(df[,-1]), 2, function(x) sum(df$Values[x]))
> data.frame(Count, Sum)
  Count Sum
A     2   4
B     1   2
C     2   7

Answer 2

An alternative using tidyr and dplyr :

library(tidyr)
library(dplyr)
df %>% gather(id, vals, -Values) %>% group_by(id) %>%
        summarise(Count = sum(vals=="Y"), 
                  Sum = sum(Values[vals=="Y"]))
#      id Count   Sum
#  (fctr) (int) (int)
#1      A     2     4
#2      B     1     2
#3      C     2     7

Data

df <- structure(list(Values = 1:4, A = structure(c(2L, 1L, 2L, 1L), .Label = c("", 
"Y"), class = "factor"), B = structure(c(1L, 2L, 1L, 1L), .Label = c("", 
"Y"), class = "factor"), C = structure(c(1L, 1L, 2L, 2L), .Label = c("", 
"Y"), class = "factor")), .Names = c("Values", "A", "B", "C"), class = "data.frame", row.names = c(NA, 
-4L))

Answer 3

Here is a data.table approach. Convert the 'data.frame' to 'data.table' ( setDT(df1) ), melt to 'long' format, grouped by 'id', get the sum of 'value' that are "Y" to get the 'Count', subset the 'Values' that corresponds to "Y" element in 'value', sum it to get the "Sum".

library(data.table)
melt(setDT(df1), id.var="Values", variable.name="id")[, {
           i1 <- value == "Y"
           .(Count = sum(i1), Sum = sum(Values[i1]))
           } ,  by = id]
#   id Count Sum
#1:  A     2   4
#2:  B     1   2
#3:  C     2   7

Answer 4

Sometimes it's easiest to just build a new data.frame from calculations on the old one:

# read in data
df <- read.table(text = 'Values   A B C
                         1        Y N N 
                         2        N Y N
                         3        Y N Y 
                         4        N N Y', header = TRUE)

data.frame(Count = colSums(df[,-1] == 'Y'),    # count of "Y"s in each column
           # sum of Values column where A/B/C is "Y"
           Sum = sapply(df[,-1], function(x){sum(df$Values[x == 'Y'])}))

#   Count Sum
# A     2   4
# B     1   2
# C     2   7