Re: does the COUNT command count missing values ?

[Ben Pfaff]

On Tue, Jan 20, 2015 at 11:28:56AM +0100, ftr wrote:
> 
> On 20/01/2015 07:01, John Darrington wrote:
> >On Mon, Jan 19, 2015 at 11:08:32PM +0100, ftr wrote:
> >      I ask this question as , to my surprise, here the COUNT command
> >      does count missing values.
> >      missing values v125 to v130 (-9).
> >      count missval = v125 to v130 (-9).
> >      results in 0 cases for missval.
> >      So I reset the missing to a valid code.
> >      Recode v125 to v130 (-9=9).
> >      count missval = v125 to v130 (9).
> >      results in about 66,000 cases without any missings (i.e. values
> >      of 9) and in 18,000 with at least one missing.
> >      This looks bizarre.
> >      Any ideas for these varying results ?
> >
> >The manual says this:
> >
> >    User-missing values of test variables are treated just like any other
> >values.  They are *not* treated as system-missing values.  User-missing
> >values that are criterion values or inside ranges of criterion values
> >are counted as any other values.  However (for numeric variables),
> >keyword `MISSING' may be used to refer to all system- and user-missing
> >values.
> >
> >
> >Are you finding that pspp is doing something different?
> I think so.
> If I understand the manual well the command should count a user-missing
> value like it counts valid values. This is what I regularly do. But in my
> case it does NOT count the user-missing value (although it is present). Only
> if I recode the missing into a valid value, in my case 9, the command works
> as intended and counts.

It does sound like a bug.  The bug might be in the code (because it does
not do what the documentation says), or it might be in the documentation
(if PSPP does the same thing as SPSS, but the documentation is wrong).

Can anyone verify SPSS's behavior here?