Re: Who "invented" the null hypthesis?


Matt Dz. wrote:
"Luis A. Afonso" wrote:

*** I do not like and I gave the raison. I repeat:
__Hypotheses tests say anything about H0 acceptance_____Paragraph.
__My opinion based on sound raisons is that I presented. (I do not intend to change the world , my job is merely to understand it, including Probability and Statistics).
__If you have different opinion, wonderful. If you think that an inappropriate term (acceptance interval versus no rejecting interval) is harmless, well, I cannot agree.

Actually, both "acceptance interval" and "no rejecting interval" are
inappropriate terms. I suggested the term "acceptance region", as it
takes into account that the parameter space is not necessarily a real
line, it is already widely accepted (and used) by the statistics
academia, and since the mistakes mentioned seem unlikely to be made I do
not consider it harmful.

However, I am not saying you are not entitled to having your own
opinion, even though I do not agree with your way of reasoning :-)

I'll leave this part of Afonso's muddle to sci.stat.math, to spare the
other two
groups of our misery in this group because of Afonso's presence.

It is more than just Afonso's opinion, it is Afonso's WRONG opinion,
which
is counter to every known definition of terms on the subject of
hypothesis
testing, that makes his statements "harmful".

The incessant NOISE Afonso's is making makes it DOUBLY harmful, not
only to the subject matter, but to the readership of sci.stat.math who
can
hardly conduct any serious discussion on the simplest subjects of
statistics or probability without Luis A. Afonso spread his ERRORS and
NOISE into dozens of threads, including the ones he created derogatory
and obscene subjects, such as "*** you, Bob" just to continue his part
of the pollution campaign.


I wrote:

*** 5% Significance Level = 95% Confidence Level ***

Well
You said that I was confused about these two terms.

Matt was correct about Afonso's confusion, and ERRORS from
the one single statement above.

I made the statement above, in the conviction that you did not understand = as EQUAL, but something as EQUIVALENT.

Even "something as EQUIVALENT" is wrong. That is what I am trying to
show - that one should NOT link these terms without specifying
underlying principles which are necessary (like the APPROPRIATE
corresponding C.I.), otherwise this in an incorrect approach - for the
reasons I have specified in the previous post.

That is:
- we can NOT say that for all hypothesis tests and confidence intervals,
even for the same parameter, \alpha significance level is equivalent to
(1-\alpha) confidence level
... as there EXISTS test with (\alpha) significance level for which the
corresponding appropriate confidence interval, for the very same
parameter, for which the confidence level is different than (1-\alpha).

I hope it is clear now.

It should be clear to most readers except to Afonso. An added faux
pas in Afonso's statement, in line with what Matt had pointed out
before, that in hypothesis testing, it is the "rejection region", which
is not necessarily an "interval". A two-tailed test would have a
rejection region consisting of the union of two "intervals" in the
parameter space. That's a fine distinction -- which is why I only
called it a faux pas on Afonso's part, about confidence intervals.

Afonso's serious ERROR about Confidence Intervals is his
unawareness that a Confidence Interval is the "single" realization
of a Random Interval.


// take the examples from my previous post, for instance

I was in error: I overestimated your capacity to understand I tried to say. In this particular, as I am a polite person,
> I do apology, like you said to do.

Please excuzzzzze me for splattering my evening drink onto my
keyboard and monitor on Luis A. Afonso's self-characterization
of a "polite person", and his overestimation of Matt's capacity to
understand. LOL!

Unfortunately, you have not shown this "politeness" in your posts.
However, I do take into account that I also might have been stern in
times. The group is already full of name-calling and ad hominem
arguments, therefore I think that keeping it clean from that would be a
good idea, and I will try to refrain from pointing out the spelling
mistakes (I did not realize you are Portuguese and I apologize if you
took some criticism personally), and I hope you could try not to post
insulting comments about choose_any_person (and that is not necessarily
only me) ;-) Honestly, posts like "THE BIG BANG (OR the pyramidal
BLUNDER)" do not contribute to this newsgroup.

Nevertheless - I ask you - simply ask - to try to posting in a correct
thread (i.e. use the tree/hierarchical structure of the NGs), and
refrain from insults/name calling/ad hominem arguments and posts with
inaccurate information (as it may be simply misleading).

A very well-formulated and very polite request of Luis A. Afonso.

Unfortunately, the probability that he'll heed the request is of the
order of 1/(1 googol ) if not 1/(1 googolplex), not to be confused
with google.


(thank you by the *advise*. If you had something to write in Portuguese I shall pay back the favour, free).

You are welcome; I will remember about the Portuguese :-)

Regards,
Matt Dz.

-- Reef Fish Bob.

.