Date: Fri, 9 Sep 2005 20:15:09 -0400
Reply-To: Richard Ristow <email@example.com>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: Richard Ristow <firstname.lastname@example.org>
Subject: Re: Explaining output of a do if with missing values
Content-Type: text/plain; charset="us-ascii"; format=flowed
At 09:59 AM 9/9/2005, Martin Sherman wrote:
>>I don't understand why if Race eq 3 and sex is missing then racesex
>>eq 5, but when sex eq 1 and race is missing then racesex eq missing
>>[in the following code, in which indents have been added for clarity]:
>>. do if (race eq 1 and sex eq 1).
>>. compute racesex=1.
>>. else if (race eq 1 and sex eq 2).
>>. compute racesex=2.
>>. else if (race eq 2 and sex eq 1).
>>. compute racesex=3.
>>. else if (race eq 2 and sex eq 2).
>>. compute racesex=4.
>>. compute racesex=5.
>>. end if.
>I guess that the cases with race = 3 are passed along for later
>processing, even when sex is missing.
Yes, as documented. In this case, "race eq 2" tests 'false', as does
"race eq 2". In SPSS logical calculations, "FALSE and MISSING"
evaluates 'false', which is logically reasonable(*). If race is 3, all
the tests will test 'false'; the ELSE clause will be reached, and
'racesex' assigned value 5.
>The cases with race = 1 or 2 and sex missing all fail in the earlier
>DO IF/ ELSE IF statements.
Again, yes. If sex is missing, "sex eq B" tests 'missing', and no
combination "race eq A and sex eq B" can ever test 'true'; see the
table below. If race is 1, the first test returns 'missing' and the
rest of the DO IF construct is skipped. If race is 2, the first two
tests return 'false', the third returns 'missing', and the rest of the
DO IF construct is again skipped. In neither case will any of the
COMPUTE statements be executed.
Now, what do you want to happen? A simple improvement is,
do if (race eq 1 and sex eq 1).
Then, racesex has value 9 if the DO IF is skipped because a test
Or, you can specify the outcome for different 'missing' combinations of
race and sex by using the MISSING function in tests. But the logic for
that depends on what you're wanting to do.
From the syntax manual:
Table 6 Outcome for AND and OR combinations
Combinations with AND:
true AND true = true
true AND false = false
false AND false = false
true AND missing = missing
missing AND missing = missing
false AND missing = false*
Combinations with OR:
true OR true = true
true OR false = true
false OR false = false
true OR missing = true*
missing OR missing = missing
false OR missing = missing
* Expressions where SPSS can evaluate the outcome with incomplete
information. See "Missing Values in Logical Expressions".