missing() not working on strings

[Akiva Liberman]

Recently, the missing() function no longer identifies empty strings for me as it had earlier. Am I the only one? I think I first noticed this a few weeks ago, but unfortunately I didn't write it down. (I updated stata moments ago just in case, but that didn't help.)

E.g., I have a string variable for social security numbers (without any dashes) called ssn. I can see empty strings in a browse window (as periods). Here are a few commands and stata's responses, copied from the output window:

. replace ssn = trim(ssn) // just to be safe
(0 real changes made)

. count if missing(ssn)
0

. count if ssn=="" // and this isn't working either
0

. destring(ssn), replace
ssn: all characters numeric; replaced as long
(170 missing values generated)

. count if missing(ssn)
170

Of course, this example is easy to work around, because I can destring; that option isn't available for many or most strings.
Any help would be appreciated!

Akiva

[Joseph Coveney]

I can see empty strings in a browse window (as periods).

A lone decimal point in a numeric variable's observation is Stata's way of indicating a missing value, but a period (full stop) in an observation of a string variable is not missing. It's a character.

Any help would be appreciated!

You can do something like the following.

replace ssn = "" if strtrim(ssn) == "."

[Akiva Liberman]

Thanks very much! That worked. I wonder how those periods got there, but that is for another day.