Re: Flip command and string vars

Posted by David Marso on Nov 19, 2013; 4:55pm
URL: http://spssx-discussion.165.s1.nabble.com/Flip-command-and-string-vars-tp5723152p5723158.html

In all likelihood because FLIP is a brain-dead procedure and all the smart people were on vacation when they had the design meetings (FLIP is certainly pre 1990).  
Hint Gene:  Both VARSTOCASES and CASESTOVARS respect strings.
Furthermore they can use mixed length strings (setting the result to maximum).
In response to many angry callers and faxes, I created a 'workaround' in the early 90's using XSAVE in a loop followed by a VECTOR/AGGREGATE combo.  IIRC it was a pain because VECTOR also requires same size strings. To make it flexible required building a new set of variables with a known max-size and tossing those into the XSAVE loop.
That is now happily long obsolete since VARSTOCASES and CASESTOVARS (circa 2000 or so).  

DATA LIST FREE / a b (2A1) c d z (3A6).
BEGIN DATA
a b c d z
z d c b a
END DATA.

VARSTOCASES /ID=id /MAKE stack FROM a TO z /INDEX=oldvar(stack) .
ALTER TYPE  id (a10).
COMPUTE id=CONCAT("Case_",LTRIM(id)).
SORT CASES BY oldvar id.
CASESTOVARS /ID=oldvar/INDEX=id /GROUPBY=VARIABLE.

Maguin, Eugene wrote

Given a dataset of only string vars of the same dimension, e.g., a5, I'm curious as to why the Flip command returns sysmis for the rearranged dataset. I understand that this is a very specific situation and maybe that's a good enough reason. And I understand why flip won't work for a mixed dataset of either strings and numeric or strings of varying dimensions.
Gene Maguin