Dear List members,
it seems to me that there are several members having long
experience with or a deep understanding of the 2X2 test situation.
Could you please advise me in the below issue:
Having the unconditional situation with two independent samples
we sometimes apply the procedures described e.g. by Agresti:
calculate alpha_obs for several probabilities and take afterwards
the supremum over all probabilities (supposing a smooth function).
This procedure works well (at least theoretically), but is a priori
conservative.
However, I do not really know, how conservative this procedure works, i.e.
do not know, whether the supremum is a good approximation (does it strongly
depend on the probability values?).
Further question is, whether this procedure leads to better results
than the application of the (conditional) Fisher test.
Is there any analytical or simulational work which could give me an
at least rough estimate about the goodness of the above procedure?
TIA for your help
Robert N�meth
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
