Data Transformation

[Riya]

Hey SPSS Listers,

Is there any command similar to ANY statement for multiple AND conditions in SPSS? I frequently use ANY statement when it requires selecting records on the basis of multiple OR condition.
For example : Select If ANY(ID,1,2,3).

I am curious to know any statement which allows to subset data on the basis of multiple AND condition.

For example : If ID = 1 AND VAR1 = 2  AND VAR2 = 3 AND VAR3 = 4.

Can the above command be shortened?

[David Marso]

Riya,
Notice that IF ANY(ID,1,2,3) is shorthand for
IF ID=1 OR ID=2 OR ID=3.
See there is a single source of variation and a constant.

Your requested AND has 2 sources of variation so how would that work syntactically?
ID = 1 AND VAR1 = 2  AND VAR2 = 3 AND VAR3 = 4

Maybe something of the form.
COMPUTE ALL4=SUM(ID=1,V1=2,V2=3,V3=4 ) EQ 4.
data list free/ id v1 to v3 .
begin data
1 2 3 4
2 1 2 3
3 4 3 2
4 2 3 4
end data.
COMPUTE ALL4=SUM(ID=1,V1=2,V2=3,V3=4 ) EQ 4.
LIST.

    id     v1     v2     v3   ALL4

  1.00   2.00   3.00   4.00   1.00
  2.00   1.00   2.00   3.00    .00
  3.00   4.00   3.00   2.00    .00
  4.00   2.00   3.00   4.00    .00


Number of cases read:  4    Number of cases listed:  4

Riya wrote

Hey SPSS Listers,

Is there any command similar to ANY statement for multiple AND conditions in SPSS? I frequently use ANY statement when it requires selecting records on the basis of multiple OR condition.
For example : Select If ANY(ID,1,2,3).

I am curious to know any statement which allows to subset data on the basis of multiple AND condition.

For example : If ID = 1 AND VAR1 = 2  AND VAR2 = 3 AND VAR3 = 4.

Can the above command be shortened?

[Ujjawal]

Hi Riya,

Along the lines of logic that David aka Mr Tech God suggested :

COMPUTE Condition = MEAN(ID=1,VAR1=2,VAR2=3,VAR3=4 )= 1.
EXECUTE.

The above mean statement will always return 1.

[David Marso]

Beautiful Ujjawal!!!
Much nicer solution since we don't need to know the number of elements
(What was I thinking -BF-:  Bruce must be taking the day off ;-)  )!
--
BTW: Please don't refer to me as Tech God ;-(
Being a student of history, myself and whatever deities that may or may not exist
don't exactly constitute a mutual admiration society ;-)
I know you mean well but ...
--

Ujjawal wrote

Hi Riya,

Along the lines of logic that David aka Mr Tech God suggested :

COMPUTE Condition = MEAN(ID=1,VAR1=2,VAR2=3,VAR3=4 )= 1.
EXECUTE.

It's always equal to 1.

[Ujjawal]

Thank you for your lovely words..This really means a lot to me..
I think *SPSS Guru* better describes you ;)

[Bruce Weaver]

I did look at it earlier, and was also thinking about MEAN, as it happens.  ;-) But I was thinking that in the event of possible missing data (and depending on what one wants to do about it), perhaps it should be:

COMPUTE Condition = MEAN.4(ID EQ 1,VAR1 EQ 2,VAR2 EQ 3,VAR3 EQ 4) EQ 1.

Like David's first version, this requires one to know the number of variables.

Notice too that I followed Art K's advice, which is to reserve = for assignment of value to variable, and to use EQ, NE, LE, GE, GT or LT when testing conditions.

David Marso wrote

Beautiful Ujjawal!!!
Much nicer solution since we don't need to know the number of elements
(What was I thinking -BF-:  Bruce must be taking the day off ;-)  )!
--
BTW: Please don't refer to me as Tech God ;-(
Being a student of history, myself and whatever deities that may or may not exist
don't exactly constitute a mutual admiration society ;-)
I know you mean well but ...
--

Ujjawal wrote

Hi Riya,

Along the lines of logic that David aka Mr Tech God suggested :

COMPUTE Condition = MEAN(ID=1,VAR1=2,VAR2=3,VAR3=4 )= 1.
EXECUTE.

It's always equal to 1.

[Ujjawal]

@ Bruce : Thanks for pointing out limitation of my suggested formula :)) It made me think of writing another formula that doesn't requires one to know the number of variables.

COMPUTE  Condition = SUM (MEAN(ID EQ 1,VAR1 EQ 2,VAR2 EQ 3,VAR3 EQ 4) EQ 1, NMISS(ID EQ 1,VAR1 EQ 2,VAR2 EQ 3,VAR3 EQ 4) EQ 0) EQ 2.
EXECUTE .

Note : I am unsure whether riya would like the above formula as it doesn't meet the purpose of eliminating AND ;))

[Bruce Weaver]

Ujjawal, note that you don't need the equality tests in your NMISS.  I.e., you could shorten it to:

COMPUTE  Condition =
  SUM (MEAN(ID EQ 1,VAR1 EQ 2,VAR2 EQ 3,VAR3 EQ 4) EQ 1, NMISS(ID,VAR1,VAR23,VAR3) EQ 0) EQ 2.

But as you suggest, this is not really gaining anything over Riya's original approach:

COMPUTE Condition =
  NMISS(ID,VAR1,VAR2,VAR3) EQ 0 and
  (ID EQ 1) and (Var1 EQ 2) and (var2 EQ 3) and (var3 EQ 4).

In fact, her original method uses fewer keystrokes, and is probably more transparent for most users.

If one needed to do something like this for a LOT of variables, a VECTOR & LOOP method could be used, with an early exit from the LOOP as soon as the condition became true.  It would not surprise me if there is an example of something like that in the archives of this list.

Cheers,
Bruce

Ujjawal wrote

@ Bruce : Thanks for pointing out limitation of my suggested formula :)) It made me think of writing another formula that doesn't requires one to know the number of variables.

COMPUTE  Condition = SUM (MEAN(ID EQ 1,VAR1 EQ 2,VAR2 EQ 3,VAR3 EQ 4) EQ 1, NMISS(ID EQ 1,VAR1 EQ 2,VAR2 EQ 3,VAR3 EQ 4) EQ 0) EQ 2.
EXECUTE .

Note : I am unsure whether riya would like the above formula as it doesn't meet the purpose of eliminating AND ;))

[Riya]

Thanks David, Bruce and Ujjawal for your efforts. My data doesn't have a lot of variables so i think it's better to go with the original AND code. But i really liked the MEAN and SUM approach :)