If a value is a dot, is it a missing?

[Sergiy Radyakin]

Dear All,

this is probably a question to the developers, but I wanted to double-check with the community before reporting.

So: if something is equal to a dot in Stata, is it a missing?

Or in other words, if X==. does this always result in missing(X)==1 or are there any exceptions? (according to help missing there are none).

Thank you, Sergiy Radyakin

Stata/MP 13.0, Windows

[Nick Cox]

I can't think of a reason to doubt this. But as you are aware,. although missing(.) yields 1 the converse isn't necessarily true. missing() yields 1 for any argument that is missing, numeric or string.

. di missing(.)
1

. di missing(.a)
1

. di missing(.z)
1

. di missing("")
1

Why ask?

[Nick Cox]

[answer posted in wrong thread]

[William Lisowski]

Let me add two cases where values present as a dot but are not missing.

. di missing(".")
0

. set obs 1
number of observations (_N) was 0, now 1

. gen x = 42

. label define x 42 "."

. label values x x

. list x in 1

     +---+
     | x |
     |---|
  1. | . |
     +---+

. di missing(x[1])
0

[Clyde Schechter]

Well, OK. But in the example in #4, x[1] isn't a dot. It's a 42 wearing a dot costume. I know that William Lisowski knows that. But there are lots of posts on this Forum relating to problems that stem from users who confuse the way a variable is displayed with what its actual internal content is. This can arise with value labels, as here, or with display formatting.

So, had I written #4, I would have presented it more as "Not everything that looks like a dot is missing, but everything that is a dot is."

[Sergiy Radyakin]

Dear Nick, William and Clyde, thank you for responding.
Your answers have reinforced my expectation that if a value is equal to a dot it must be also missing in the sense of the function missing(). Since I observe the opposite, I will be reporting this as a bug to Stata developers:

NAME1 is a double storage type variable. The name of the variable is irrelevant. I can add that the value in NAME1[1] is not a result of a typical transformations you do in Stata, so a Stata user is unlikely to encounter the same in a regular session. But that particular value then causes consequences throughout Stata:

. misstable summarize
                                                               Obs<.
                                                +------------------------------
               |                                | Unique
      Variable |     Obs=.     Obs>.     Obs<.  | values        Min         Max
  -------------+--------------------------------+------------------------------
         NAME1 |         3                   0  |      3          .           .
     benchmark |         1                   2  |      2          1           2
  -----------------------------------------------------------------------------

In the meanwhile, I use the workaround x>=. to identify missing values instead of the function missing().

Thank you,Sergiy Radyakin

[William Lisowski]

Sergly - Could you share with us the result of

di %21x NAME1[1]
di %16H NAME1[1]

[Sergiy Radyakin]

Dear William, sure:

. di %21x NAME1[1]
+1.0000000000000X+3ff

. di %16H NAME1[1]
7fe0000000000000

. di NAME[1]==.
1

. di missing(NAME[1])
0

I have submitted the case with detailed description and a pointer to this thread to Stata's tech-support.

Best, Sergiy

[William Lisowski]

Thanks, Sergly. I will be interested in the response. The values displayed are identical those created by for example gen double x = . which of course produces a value that works as expected. These results lead me to suspect, and I think your example confirms, that using display with these formats does not guarantee an exact representation of the contents of a variable in memory. I'm guessing that's because NAME[1] is implicitly an expression and gets passed through Stata's expression evaluation before being passed to display.