CODEWITHSUNDEEP
rindexing
nicolas
nicolas

Reputation: 9805

vectorizing "if" in R

I have an array of probability over a classification 

post <- c(0.73,0.69,0.44,0.55,0.67,0.47,0.08,0.15,0.45,0.35)

and I want to retrieve the predicted class. right now I use 

predicted <- function(post) { 
               function(threshold) {plyr::aaply(post, 1, 
                   function(x) {if(x >= threshold) '+' else '-'})}}

but that seems like something R would have a syntax for.

Is there some indexing expression that would be more direct ?

Upvotes: 3

Views: 166

Answers (2)

Ben Bolker
Ben Bolker

Reputation: 226097

As @joran suggests:

predicted <- function(post)
   function(threshold) 
      ifelse(post>threshold,"+","-")

I find the nestedness of the functions a little confusing.

ifelse(post>threshold,"+","-")

seems sufficiently simple that you might not even need to package it in a function.

Or you could use 

predicted <- function(post,threshold=0.5,alt=c("+","-"))
      ifelse(post>threshold,alt[1],alt[2])

Alternatively

predicted <- function(post,threshold=0.5,alt=c("+","-"))
   alt[1+(post<=threshold)]

would probably be marginally faster (post>threshold gives a logical vector, which is coerced to 0/1 when added to 1, resulting in 1 for "below" and 2 for "above"). Or you could reverse the order of alt, as @DWin does in his answer.

Upvotes: 6

IRTFM
IRTFM

Reputation: 263301

pred <- c("-", "+")[1+(post > 0.5)]

Upvotes: 7

Related Questions