Re: A * Titans * fight II: Jack Tomsky versus R. A. Fisher

In order to clarify our opposite positions concerning
this point it was interesting to us, the Readers, you
answer the question:

__1__Under what conditions you think the Null
Hypotheses, when applied to a CONTINUOUS PARAMETER
HYPOTHESES TEST, _____can be accepted because it´s
true?




The null hypothesis should always be accepted whenever the sample is in the acceptance region.




My position is

__1__The existence of a finite C.I. defining the
complementary of the critical region (acceptance
interval) in the case that CONTINUOUS PARAMETERS are
involved, forbade us to think that to accepted the
NULL HYPOTHESES: there are n means to ascribe ZERO to
the parameter (or Zero to the parameters difference
if it’s the case).
__2__In the case of the test of the difference
between two discontinuous parameters the nullity can
be achieved but it´s a very rare situation and at
growing sample sizes practically impossible.
Furthermore, when one is testing the Population
Binomial Proportions one has to draw conclusions from
the Probability Function, never from any approximate
procedure (Normal).




That's why no one should use Afonso's equations which lead to results that he now disowns. It is better to use the exact test given in

Tomsky, J.L., "Statistical comparison of two failure rates from either binomial or negative binomial sampling", presented at the Seventh SIAM Conference on Discrete Mathematics, 1994.

Jack (moderator)
.

Relevant Pages