Re: JACK TOMSKY: an HISTORIAL of nonsense

___Richard Atkins wrote

*** Date: Mar 19, 2007 6:49 AM


*** If you mean true in the sense of absolutely true, then when working with statistics (i.e. samples) clearly neither the null or alternative hypothesis can ever be stated as being true. Perhaps I have misunderstood but this does not seem to be the point - see below.***

My response

Thank you by your concern.
I am not, surely, a high level statistician. Professionally my job was to do data processing.
We are, you see, in perfect agreement concerning the impossibility to state a hypothesis (null or alternative) is true.
On contrary our job is to state whatever the circumstances:
__The Hypotheses Test is only a *tool* showing us, at a convenient level, if we have sufficient evidence to reject H0, or to keep this hypotheses_____
In consequence:
____When it posed H0: m1 = m2 in testing two parameters the equal signal is not, BY ANY MEANS, an algebraic one, but merely a convention that has a precise meaning:
__ The m1 and m2 parameters are so alike that, facing the data, and at a convenient significance level, we have not sufficient evidence they are different.
_______________
I do not find any justification in teaching differently at whatever degree.
It would avoid, manly among *the technical statistical consumers*, wrong ideas about hypotheses testing. It is sad (and even scandalous) their teachers had not the concern to enlighten them from the start.

Thank you for your time

_________licas (Luis A. Afonso)
.

Quantcast