Question

Created a data frame:

simpleDF<- structure(list(vals = c(NA, NaN, 2)), .Names = "vals", row.names = c(NA, 
-3L), class = "data.frame")


> is.na(simpleDF$vals)
[1]  TRUE  TRUE FALSE
> is.nan(simpleDF$vals)
[1] FALSE  TRUE FALSE

Now, when I change the data frame to include a string value:

simpleDF <- structure(list(vals = structure(c(NA, 2L, 1L, 3L), .Label = c("2", 
"NaN", "test"), class = "factor")), .Names = "vals", row.names = c(NA, 
-4L), class = "data.frame")

> is.na(simpleDF$vals)
[1]  TRUE FALSE FALSE FALSE
> is.nan(simpleDF$vals)
[1] FALSE FALSE FALSE FALSE

I did not understand why NaN is no longer recognized by is.nan()? Sure there would be an explanation ..

Answer 1

NaN is a value that only makes sense in numeric vector, so it gets converted to the character string "NaN" when the class of the vector it is in is (explicitly or implicitly) converted from numeric to character or factor.

vals
# [1]  NA NaN   2
as.character(vals)
# [1] NA    "NaN" "2" 
c(vals, "A")
# [1] NA    "NaN" "2"   "A"