Ted Dunning has already provided a very useful reply to this question.
I can't add to his discussion, although I'd just like to make a book
recommendation for anyone interested in these sorts of tests:
@book{ReadC88,
author={Read, T. and Cressie, N.},
title={Goodness of fit Statistics for Discrete Multivariate Data},
year = {1988},
address = {New York, NY},
publisher = {Springer-Verlag}}
This book gets into the whole controvery of Pearson's X^2 test vs the
log--likelihood ratio (as did Ted Dunning in CL). It also talks about
exact tests as alternatives to the asymptotic tests and gives some
guidance as to when you should expect certain tests to be valid and when
they might not be.
Some of the same issues are also discussed in this earlier article:
@article{CressieR84,
author = {Cressie, N. and Read, T.},
title = {Multinomial Goodness of Fit Tests},
journal = {Journal of the Royal Statistics Society Series B},
volume = {46},
year = {1984},
pages = {440-464}}
The book is of course more detailed but both are very nice sources.
Best regards,
Ted
-- * Ted Pedersen pedersen@seas.smu.edu * * http://www.seas.smu.edu/~pedersen/ * * Department of Computer Science and Engineering, * * Southern Methodist University, Dallas, TX 75275 (214) 768-3712 *