Date: Dec 22, 2012 7:42 PM
Author: Luis A. Afonso
Subject: Re: Who cares the null hypotheses (two-tail test) is really TRUE?
?Should Psychology abandon p-values and teach CI´s instead? Evidence-based reform in Statistics education?.
Fiona Fidler (2006).
I, myself, never search if a hypothesis is true or false . . . a chimerical goal I do not pursue. We should not care, at all, such a thing: is impossible and uninteresting. What we are only able to achieve is if there is (or not) sufficient evidence/plausibility to reject the Null Hypothesis giving the data. To bring the Aristotelian logic (true/false) to events subjected to chance is plainly erroneous.
The A. said, all people agree: ?particularly serious misconception associated with NHST, namely that statistical non-significance is equivalent of evidence of `no effect`. But immediately she state that ? from a low power study with a no-trivial effect size - as evidence the null hypotheses is true . . . suggest that she have the concern, about she criticise as a method?s weakness: a sufficiently large sample will invalidate this result. But this is expected, of course, this is natural: even a little difference will be evident from data, therefore leading to H0 rejection. A difference that could be, evidently, insignificant for practical purposes.
Luis A. Afonso