CODEWITHSUNDEEP
rnaninf
Mahan
Mahan

Reputation: 109

What is the best way to remove infs or NaNs in R?

I'm having some trouble with infs or NaNs in my files. What is the best way to remove them using R?

Upvotes: 0

Views: 291

Answers (2)

DaveArmstrong
DaveArmstrong

Reputation: 21757

If you're trying to get rid of NaN and Inf of -Inf, you're better off using is.finite. In the context of a data frame, you could eliminate the offending rows with:

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
  dat <- data.frame(
   x1 = c(1, 2, NA, NaN, 3, Inf),
   x2 = c(1,2,3,4, NaN, -Inf)
 )
dat
#>    x1   x2
#> 1   1    1
#> 2   2    2
#> 3  NA    3
#> 4 NaN    4
#> 5   3  NaN
#> 6 Inf -Inf
dat %>% filter(if_all(everything(), is.finite))
#>   x1 x2
#> 1  1  1
#> 2  2  2

Created on 2022-04-05 by the reprex package (v2.0.1)

Note that !is.na() and !is.nan() don't solve the whole problem:

dat %>% filter(if_all(everything(), ~!is.nan(.x)))
#>    x1   x2
#> 1   1    1
#> 2   2    2
#> 3  NA    3
#> 4 Inf -Inf
dat %>% filter(if_all(everything(), ~!is.na(.x)))
#>    x1   x2
#> 1   1    1
#> 2   2    2
#> 3 Inf -Inf

Upvotes: 3

Sean McKenzie
Sean McKenzie

Reputation: 909

You can use !is.nan(). Here is an example:

x1<-sample(1:50, 10)
x2<-sample(c(0, 5,4), 10, replace=TRUE) 
x3<-x2*0.5
x4<-x1/x2 #will generate Inf because we are dividing by 0 
x5<-x1/x3 #will generate Inf because we are dividing by 0
x6<-x5-x4 #will generate NaNs because we are subtracting Inf from Inf
x7<-x6[!is.na(x6)]#Will remove all NaN elements

Upvotes: 0

Related Questions