Can I calculate variance using dplyr::mutate?

Question

I'd like to create a new column in my data table with variance of values that are stored in some other columns that I have in my table. I thought the mutate function from dplyr package would be the best solution for me, but unfortunately it does not seem to work.

Here is the code that I'm using:

require(dplyr)
CPI <- read.csv("CPI.csv", sep = ";")
CPI <- tbl_df(CPI)    
CPI <- mutate(CPI, variance = var(CPI.1:CPI.6, na.rm = T))

and this is the error that I get in return:

Error in c(1.3, 1.4, 2.9, 3.1, NA, NA, NA, NA, NA, NA, NA, NA):c(NA, 1.2253333335,  : 
  NA/NaN argument

Is it possible to calculate the variance using mutate? Should I use other solution?

Answer 1

I would be surprised if CPI.1:CPI.6 evaluated as you expect. Why not just take your original data.frame and do this:

m <- as.matrix(subset(CPI, select=CPI.1:CPI.6))
apply(m, 1, var, na.rm=TRUE)

Or use the matrixStats package for speed and clarity:

rowVars(m)

Answer 2

If you need the variance for each row, an option will be to use do with rowwise()

library(dplyr)
nm1 <- paste('CPI', 1:6, sep='.')
CPI %>% 
    rowwise() %>% 
    do(data.frame(., variance=var(unlist(.[nm1]), na.rm=TRUE)))

data

set.seed(25)
CPI <- as.data.frame(matrix(sample(c(NA,1:40), 7*20, replace=TRUE), 
           ncol=7, dimnames=list(NULL, paste('CPI', 1:7, sep="."))))

Answer 3

It looks like you are need to perform the operation row-by-row. You can try this

require(dplyr)
CPI <- read.csv("CPI.csv", sep = ";")
CPI <- tbl_df(CPI)    
CPI <- group_by(CPI, 1:nrow(CPI))    
CPI <- mutate(CPI, variance = var(CPI.1:CPI.6, na.rm = T))

Or are you trying to do anything else?