Re: Bob Ling´s ignorance is clear
- From: "Old Mac User" <chendrixstats@xxxxxxxxx>
- Date: 14 Jan 2007 17:04:29 -0800
Luis...
This will be brief because I'm ready to stop for this day.
I have no clue how you came to this... you wrote: "His insistence is a
clear mark how he thinks that the *normal approximation* is the best we
can do in order to compare two INDEPENDENT proportions."
To the contrary, I have repeatedly suggested with emphasis that
Fisher's Exact Test is most appropriate for comparing two independent
proportions. In doing this I've noted that the classic and
well-established Chi-sq method (with Yates correction) produces "probs"
that are very close to Fisher's Exact Test. When Yates correlation is
applied appropriately, Chi-sq offers an extremely good approximation to
Fisher's Exact Test.
There are other variations on this theme. For instance, Resampling can
be used to compare two independent proportions. Resampling yields
"probs" that converge to Fisher's Exact Test probs.
Not once have I suggested using the Normal approximation to a binomial.
To the contrary, you are the one who has been proposing the normal
approximation.
For reasons that Reef Fish explained with clarity, the classic "null
hypothesis" establishes a "prob" for the case in which two sources of
data have the same average, fraction, proportion, or rate. This
affords us a clear and concise way of establishing the "Type I Error
Rate" and a way of making comparisons among competing methods of
comparing averages, etc.
In the meantime, I have repeatedly asked that you explain "your method"
with numbers, and do that completely with nothing left out. This, for
the sake of showing all of us exactly what you are suggesting. This is
commonly called "a sample calculation". I have not yet seen such a
sample calculation. OMU
"Luis A. Afonso" wrote:
As long as OMU uses dilactoris tactics for not to solve the problems I posed all kind of discussions, are useless. The trick has clearly that goal.
The 2 problems are rigorously posed (Posted: Jan 14, 2007 3:04 AM) in respect to an OMU ´s asking for NUMERICAL DATA, H0 and respective Confidence Levels.
IS HE ABLE TO DO SOLVE THEM?
The following text is a complete NONSENSE:
*** My attempts at explaining certain basic facts to Afonso failed miserably. He still doesn't understand the basic concept of beginning with the conjecture that two sources of data may have the same true average or the same frequency of outcomes and working forward from there. Anything I might do to explain this in new and different ways would be lost in the noise. Afonso's free education came to an abrupt halt on 1/1/07. Maybe someone else wants to pick up the flag and try. And try. And try. And try. OMU***
1) BY THE HELLWHERE I DENNIED THAT two sources of data could have the SAME TRUE AVERAGE? I simply can assert if THAT is admissible (from the observed data) or unlikely.
2) Clearly OMU does not want to enter on the discussion REALM. Obviously he uses personal (arbitrary) prejudicing (and rhetoric) *noise* in order to avoid (incompetence or bad faith?) to go further.
His insistence is a clear mark how he thinks that the *normal approximation* is the best we can do in order to compare two INDEPENDENT proportions.
______licas (Luis A. Afonso)
.
- References:
- Re: Bob Ling´s ignorance is clear
- From: \"Luis A. Afonso\"
- Re: Bob Ling´s ignorance is clear
- Prev by Date: Re: why is linear regression with L1 much better in this case?
- Next by Date: Re: Best modeling approach for dependent variable with 10 ordered groups?
- Previous by thread: Re: Bob Ling´s ignorance is clear
- Next by thread: Re: statistics versus biostatistics
- Index(es):
Relevant Pages
|
