CODEWITHSUNDEEP
r
Zach
Zach

Reputation: 30301

Function to identify problematic datatypes

I just spent several hours debugging some R code, only to discover that the error was due to an Inf that had sneaked in during my calculations. I had checked for NA, but hadn't thought to check for Inf.

I wrote the following function to help prevent this situation in the future:

is.bad <- function(x){
  is.na(x) | is.nan(x) | is.infinite(x)
}

>     is.bad(c(NA, NaN, Inf, -Inf, 0, 1, 1000, 1e6))
[1]  TRUE  TRUE  TRUE  TRUE FALSE FALSE FALSE FALSE

Are there any other special data types in R I should be aware of?

Upvotes: 1

Views: 71

Answers (3)

Ben Bolker
Ben Bolker

Reputation: 226097

!is.finite(x) is equivalent to your is.bad(x); it detects Inf (and -Inf), NA, and NaN values.

Upvotes: 4

Se&#241;or O
Se&#241;or O

Reputation: 17412

It depends a bit on what you're trying to do. I had a similar problem recently when I was creating a linear model for a bunch of data subsets. What worked for me to get coefficients when they were possible, and an NA instead of a program stopping error message was:

Coef <- ifelse(is.numeric(try(coef(lm(y~x, data=DF)), silent=T)), coef(lm(y~x+0, DF)), NA)

So I would get a vector

 [1] 2.3 4.3 5.4 6.2 2.8 NA 3.2

Properly spaced, instead of an error message.

Upvotes: 1

Stephan Kolassa
Stephan Kolassa

Reputation: 8267

Depending on what you are doing, is.null() could possibly at some point also save you a lot of time. Hard to say in general.

Upvotes: 1

Related Questions