[prev in list] [next in list] [prev in thread] [next in thread] 

List:       sas-l
Subject:    Re: Exact Test, 3x4 untenable?
From:       John Whittington <medisci () POWERNET ! COM>
Date:       1999-01-29 0:25:49
[Download RAW message or body]

At 03:38 28/01/99 GMT, Adam Sundor wrote:

>Using a 2x4 table, the Exact Measures ran in seconds on a
>unix HP machine, or seemed so.
>I thought I would add another column for the relatively few "Don't Know"
>answers to see what the results would be.  Not necessary
>and does not seem logical for the analysis, but I was curious.
>The cpu time went up to almost 50%, where I was dueling it out
>with a matlab job, but when the memory usage climbed to 50 MB
>I suspected a problem.

Adam, I fear that all you are seeing in the reality of the present state of
SAS 'exact' test realisation.  Using just tiny cell sizes (average cell
counts of about 10, random), on my PC (with 64 MB RAM), a 2x4 table takes
0.2 seconds a 3x4 one takes about 12 seconds, and I gave up at about 15
minutes waiting for a 4x4 - but I can do a 4x4 in about 58 seconds if the
average cell count is reduced to about 5!  Even with the 3x4, if average
cell size increases much above 10, the execution time becomes 'ridiculous'.

If there's any cosolation at all, it's that SAS does warm me in the log, and
give me a 'bail out' option!! :

WARNING: Computing exact p-values for this problem may require much time and
         memory. Press the system interrupt key to terminate exact
         computations.

Regards,

John

----------------------------------------------------------------
Dr John Whittington,        Voice:    +44 (0) 1296 730225
Mediscience Services        Fax:      +44 (0) 1296 738893
Twyford Manor, Twyford,     E-mail:   medisci@powernet.com
Buckingham  MK18 4EL, UK              mediscience@compuserve.com
----------------------------------------------------------------

[prev in list] [next in list] [prev in thread] [next in thread]