topic Re: iml avoid error with empty matrix in SAS/IML Software and Matrix Computations

iml avoid error with empty matrix

acordes — Fri, 03 Mar 2017 20:19:38 GMT

I run into an error while assigning a scalar with the loc function.

Everything works fine until I do not meet the condition in the loc.

Break_even = MIN(loc(result > 0));

This returns the month when the project breaks even. But if the project fails to do so, the code stops executing because the matrix Break_even isn't set.

I use this inside a Do Loop and I have tried to control this situation with If-then.

without success so far.

How do I handle situations like this?

Thank you for your help, bye, Arne

Re: iml avoid error with empty matrix

Rick_SAS — Fri, 03 Mar 2017 21:07:30 GMT

using the LOC function as an index without checking that the return value is nonempty is jokingly called "using a naked loc". Read about the naked loc and how to test for the condition.

The typically usage is

idx = loc(result > 0);

if ncol(idx) > 0 then

Break_even = min(idx);

else

/* handle the case where no result is positive */

How you handle the empty result is up to you. You might assign a missing value, abort the program, assign a default value, and so forth.

Re: iml avoid error with empty matrix

acordes — Fri, 03 Mar 2017 22:03:47 GMT

Thank you very much.

I have applied your proposal

BUT I still get the error message

ERROR: ELSE does not follow IF statement at line=...

ERROR: END does not occur within DO group at line...

As I had mentioned, I use it within a Do Loop and making use of a SUBMIT statement I call SGPLOT to produce 64 graphs according to my different scenerios. It works perfectly except this abort when the there's no break even 🙂

Re: iml avoid error with empty matrix

Rick_SAS — Fri, 03 Mar 2017 22:09:53 GMT

The errors indicate that you did not implement the IF-THEN/ELSE condition correctly. If you post the relevant portion of your log, someone will be able to find the mistake.

Re: iml avoid error with empty matrix

Ksharp — Sat, 04 Mar 2017 03:20:15 GMT

OR Try any()

if any(result > 0) then Break_even = MIN(loc(result > 0));

Re: iml avoid error with empty matrix

acordes — Mon, 06 Mar 2017 11:50:00 GMT

Thank you.

It works except when I run an if-then statement within a do-loop.

Do I have to obey something special when I place an if-then logic within a do-loop?

The do-loop executes fine without if-then and the same is true for the inverse.

But both togehther gives error.

Re: iml avoid error with empty matrix

Rick_SAS — Mon, 06 Mar 2017 12:17:47 GMT

No, the syntax for an IF-THEN statement inside a DO loop is the same as outside. You might have missing or mismatched DO-END statements.

Please post the portion of the SAS log that gives the source and error messages.

Re: iml avoid error with empty matrix

IanWakeling — Tue, 07 Mar 2017 10:37:40 GMT

If you know the size of the matrix result, or can easily work it out, then another way might be to use CHOOSE. Here I am assuming result is a row vector with 12 elements, one for each month.

   b = min( choose(result>0, 1:12, . ));

I was thinking I would get a missing value returned when all the elements of result are negative, however I get the largest representable floating point. This is in 13.1

Re: iml avoid error with empty matrix

Rick_SAS — Tue, 07 Mar 2017 10:51:04 GMT

@IanWakeling Since the MIN function ignores missing values, you are getting the default initialized value, as documented. See the second paragraph of the doc of the MIN function.

Re: iml avoid error with empty matrix

IanWakeling — Tue, 07 Mar 2017 11:09:47 GMT

Thanks Rick. I should have checked the docs more carefully. I think I will stop there, as adding something like

if b < constant('BIG') then...

will start to look very confusing.

Re: iml avoid error with empty matrix

Rick_SAS — Tue, 07 Mar 2017 11:23:01 GMT

An easier-to-read version might be

if all(result=.) then 
   handle_all_missing;
else
   b = min(...);