[Squeak] Newbie permutations -- on sample size

Michael Grant mwgrant2001 at yahoo.com
Wed Apr 10 03:58:40 UTC 2002

Just to add some non-Squeak-specific context on the
interesting discussion on implementation discussions
on 'permutationsDo:'

MY immediate interest in permutations is to develop
permutation testing -- sometimes the only viable
statistical testing technique for SMALL sample sizes.
In that particular case the number of permutations is
manageable and indeed as the sample size goes up one
would naturally move to resampling with replacement,
e.g., randomization tests and perhaps bootstrapping. 

In these circumstances memory usage is likely not
critical--assuming intelligent, knowledgeable usage.
On the other hand, if the application required the
evaluation of permutations for larger sample sizes,
genetics applications(?) then memory quickly could be
a problem. Implementation then seems to depend
strongly on application--given the factorial growth
with sample size. In other words, no shoe size fits

I guess my point is that if I was working with small
samples then I might value consistency in syntax over
memory efficiency. If I was working with moderate and
large samples then memory issues are paramount. So we
roll our own...its in the spirit of the beast.

In any case, both the discussion on the immediate
mechanics of my permutation question, and the ensuing
discussions have been very informative to this
newcomer. I thank you all. 

Michael Grant

Perhaps a discussion of permutations merits swiki or
tutorial space--from a number of potentially very
informative perspectives. 

Do You Yahoo!?
Yahoo! Tax Center - online filing with TurboTax

More information about the Squeak-dev mailing list