I am looking for source code, or pseudo code for an implementation of the SPRT
(or preferably the M-SPRT [multi hypothesis SPRT]) to use in some
testing software.  An example of the calculation can be found here:

http://home.clara.net/sisa/sprt.htm

the statistics version can be found here:

http://www.stats.ox.ac.uk/~steffen/bs2siMT04/si15c.pdf

The implementation I am looking for should take 2-3 parameters

ratio of results for set 1, (or current # positives)
ratio of results for set 2, (or current # positives)
if # of positive, then # of tests

I may also be missing another parameter here, I'm not sure.

A multi hypothesis test which can take N options and return a list of
results for each of the results against the first would be very nice
to have.


it can assume that 1 will always be >= 2
and should return the likelihood that 1 will remain > 2
the likelihood can be 0-1 probability, or a raw score as long as you
can tell me the likelihood table or calculation I can use to convert
it into a 0-1 probability.