Missing does not equal blank: your SAS jedi mind tricks won’t work here.

I inerited a piece of R code written by a former SAS programmer. It did not work as intended.

I came across some code at work where the author wanted to condition an action based on the presence of missing data. Many of the people I work with also started out as and still work as SAS programmers, so we have a habit of trying to do things in R like:

if (x == "") {do something}

rather than is.na(x). I found the bug because the code wasn’t working properly. It was supposed to do something when data was missing. I had missing data but nothing was happening as a consequence.

Let’s see the difference between the the two.

x <- c("a", "b", "", NA, "c")
y <- c(1, 2, NA, 3)

Find the missing data with x == ""

x == ""

[1] FALSE FALSE  TRUE    NA FALSE

y == ""

[1] FALSE FALSE    NA FALSE

What about is.na()?

is.na(x)

[1] FALSE FALSE FALSE  TRUE FALSE

is.na(y)

[1] FALSE FALSE  TRUE FALSE

So, x == "" is true if it is literally true. In SAS programming, we can use "" as shorthand for missing. But with R, if a vector has a value NA, the logical test returns NA. The code I was working with failed because the missing data was NA so the logical test in the code did not return TRUE.

The function is.na() is the best way to find missing values in both character and numerical vectors with one small caveat; I have seen character data where missing values were recorded as "". If you have data like that, is.na() won’t find it.