Question

I am trying to calculate some simple ratios and use R’s bracket notation to denote the baseline for the ratio.

Now I am struggling with defining a function that lets me parameterize the baseline.
I don’t want to hardcode it, since I have a couple of those. And, I really don’t understand what R is exactly doing and am very curious how to achieve the desired behavior.

Here some code based on example data:

data("singer", package = "lattice")

# this is what I want, but what currently doesn't work
my_ratio <- function(voice) {
  ddply(singer, ~ voice.part,
        transform,
        # how do I refer to the voice variable here?
        # it looks like it misunderstands it as column?
        ratio = height / mean(height[voice.part == voice]))
}

# this version works with a hardcoded voice part
my_ratio_hard <- function() {
  ddply(singer, ~ voice.part,
        transform,
        ratio = height / mean(height[voice.part == "Soprano 1"]))
}

Answer 1

How about this:

my_ratio <- function(voice) {
  my_transform <- function(x) {
    transform(x, ratio = height / mean(height[voice.part == voice]))
  }
  ddply(singer, ~voice.part, my_transform)
}

Did you really only want to scale the rows matching voice and have NaN elsewhere (that’s also what your hard-coded function does)?

A more compact version of the above:

my_ratio <- function(voice) {
  ddply(singer, ~voice.part, 
    function(x) transform(x, ratio = height/mean(height[voice.part == voice])))
}

and if you actually want to scale all the records (as your comment suggests):

my_ratio <- function(voice) {
  scale <- with(singer, mean(height[voice.part == voice]))
  ddply(singer, ~voice.part, 
    function(x) transform(x, ratio = height / scale))
}