Re: do repeat

Posted by David Marso on
URL: http://spssx-discussion.165.s1.nabble.com/do-repeat-tp5719707p5719721.html

Thank You for that Richard,
I like your mop up much better than mine.
using the Transformation program one would:
DO REPEAT v= list1 / w=list2.
DO IF (v NE 0).
COMPUTE w=1/v.
END IF.
END REPEAT.
---
I suppose my MATRIX hack is a little unorthodox but by choosing an INFINITESIMAL
term to add to the vector I don't need to explicitly check and it comes back into SPSS with a fudge factor which can not be detected in the precision available to the main system.
AND it SLAMS it all through with one line of clever code.  
MATRIX is a cranked up energizer bunny when it comes down to performance!!!
It truly excels at vector based operations but is a little slow with item by item looping.
Now, I have no idea how MATRIX can represent such very small values and avoid the zero divide infinity.
But, AFAICT it works like a charm.  The only mop up required are the HUGE lingering values which require a RECODE or other technique to remove.
If there is a downside to this I would be most interested in your analysis and counterexamples.
 

Richard Ristow wrote

At 09:43 AM 4/25/2013, David Marso wrote:

>If the numbers are all positive then you can nuke the very small fudge factor
>and remove the recode.
>If you have negative values then you will need to modify the recode (I'm
>sure you can sort that).

>RECODE ALL (1.000001 THRU HI=SYSMIS).

I think a cleaner way to do this RECODE, sidestepping having to
choose a fudge factor, is

RECODE ALL  (LO THRU 1.0 = COPY)
             (ELSE        = SYSMIS).

But actually, I think the dividing-by-zero problem is a strong reason
to do the computation in a transformation program, rather than MATRIX.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD