Question

I have a factor RFyhat which I’m looking to convert to a numeric vector. I’ve already discovered that

as.numeric(levels(RFyhat))[RFyhat]

works as desired, and I’ve played around a bit with this construction:

c(1,2,20,4,5,6,7)[RFyhat]

also works as expected (RFyhat has 7 levels).

So I understand the behavior of this construction, but I’m wondering if anyone can explain how this syntax is intended to work, or whether it is just ‘syntactic sugar’. More specifically, does [RFyhat] act as an index vector? If it does, how do factors generally behave when used as an index?

Answer 1

Yes, I believe that factors gets converted to integers when used for indexing, rather than characters or anything else.

Look at this example

> fac <- factor(letters[c(1,1,2,1,3,3,2,1)])
> vec <- c(b=1, a=2, c=3)
> vec[fac]
b b a b c c a b 
1 1 2 1 3 3 2 1 

So element 1 of fac has returned element 1 of vec, regardless of the different order of names.

Personally I’d prefer as.integer(as.character(RFyhat)) to as.numeric(levels(RFyhat))[...].