Date: Dec 20, 2012 6:49 PM
Author: Luis A. Afonso
Subject: Re: Problems with NHST? None at all . . .

Why we shouldn?t have any problems to use NHST


Because

___1) I do not try to prove that H0 (or H1) is true or false,
___2) I never perform a Nil test which only says that if data is or is not confoundable with H0. In particular H0= p is a cryptic way to say if these quantities are not discernible given the data,
___3) I know that 1-pvalue is not a measure of the probability H0 is true: pvalue < alpha only suggests that or H0 is false or perhaps I have got an extremely odd data and then I risk the probability alpha to be wrong by rejecting the Null (Type I error),
___4) I?m aware that to evaluate a practical difference D (positive or negative) I should sum it to the Null Hypotheses value: if happens a test rejection I?m sure that the parameter?s Population value is different from the null by D with 1- alpha probability.

Luis A. Afonso