在 dcast.data.table 中使用 weighted.mean 时出错

Question

I am experimenting with dcast.data.table for weighted.mean . However it throws an error for this function.

library(data.table)
dat = data.table(
  x = c(1,2,3,1,2,3,1,2,3,1,2,3,1,2,3,1,2,3), 
  y = c(4,4,4,4,4,4,5,5,5,5,5,5,6,6,6,6,6,6), 
  z = c(7:24), 
  w = c(0.1, 0.1, 0.1, 0.9, 0.9, 0.9, 0.2, 0.2, 0.2, 0.8, 0.8, 0.8, 0.3, 0.3, 0.3, 0.7, 0.7, 0.7)
  )
dcast.data.table(
  dat,
  x~y,
  fun.aggregate = weighted.mean, w = 'w',
  value.var= 'z'
)

# Error in weighted.mean.default(z, w = "w") : 
#   'x' and 'w' must have the same length

There are workarounds that suggest to use either dplyr or data.table[] but none explain why dcast doesn't work.

As @Frank points out, the fun.aggregate argument of dcast can only take functions whose output is a single value. However, I don't think that this is the issue with weighted.mean . If I don't specify weights it get valid answer

dcast.data.table(
  dat,
  x~y,
  fun.aggregate = weighted.mean, 
  value.var= 'z'
  # ,w = 'w'
)

This is also demonstrated with quantile function which gives me a valid answer when the result for each function is a single value (ie by specifying single value for probs )

dcast.data.table(
  dat,
  x~y,
  fun.aggregate = quantile, 
  value.var= 'z',
  probs = c(0.25)
)

However when it is written to output a vector for each combination, I get an error commensurate with the limitation of fun.aggregate but different from the error I get with using weighted.mean

dcast.data.table(
  dt,
  x~y,
  fun.aggregate = quantile, 
  value.var= 'z',
  probs = c(0.25,0.75)
)
# Error: Aggregating function(s) should take vector inputs and return a single value (length=1). However, function(s) returns length!=1. This value will have to be used to fill any missing combinations, and therefore must be length=1. Either override by setting the 'fill' argument explicitly or modify your function to handle this case appropriately.

It seems that dcast doesn't split up the w argument for each function and passes the entire vector to weighted.mean function. I want to understand what internally prevents this function from doing this.

Answer 1

Wath about this?

dat = data.frame(x = c(1,1,2,2),
y = c(4,4,5,5),
z = c(1,2,3,4),
w = c(1,2,1,2))

weighted.sum
reshape2::dcast(data =  dat, formula=x~y, 
fun.aggregate = function(x){mean(x*dat$w)*length(x)},
value.var= c('z'))

#weighted.mean
reshape2::dcast(data =  dat, formula=x~y, 
fun.aggregate = function(x){mean(x*dat$w)}, 
value.var= c('z'))