Re: Testing the Equality of Two Population Proportions


Jerry.Dallal@xxxxxxxxx wrote:
Reef Fish wrote:
Jerry.Dallal@xxxxxxxxx wrote:
Reef Fish wrote:
Jerry.Dallal@xxxxxxxxx wrote:


Who cares about your 1977 American Statistician paper. I had given
you ALL the REASONS. You haven't given one single reason WHY
you don't use the pooled variance given all the reasons to use it.

So, you have not given one single valid reason WHY you want to
estimate the variance of p-hat when there is only ONE p-hat,
(when Ho is true) by a variance estimate based on two separate
p-hat estimates.

-- Reef Fish Bob,


Bob,

I HAVE given a reason, REPEATEDLY! The 1977 American Statistician
paper, which you REFUSE to read, despite your online access.

No, I DO NOT have the JSTOR or other internet access to any paper or
book.
I thought you knew that the last time you referred to some paper that
way.

I find it hard to believe that, as Emeritus faculty, Clemson does not
provide you with JSTOR access through its website.

I find it even harder to believe that, as a present statistical faculty
of any
university, someone cannot explain, for the ONE, SINGLE, NARROW
special case (known to all Freshman students I taught) without
having to rely on some non-specific point in an obscure paper..

I encourage everyone to think of that NARROW issue of Ho: p1=p2.

Because that is the ONLY issue at stake. The only special case
where pooling is used for the variance. It doesn't even apply if
the Ho is p1-p2 = .1 or anything else. It has to be ZERO.

Read the above AGAIN. It is very, very narrow SINGLE case.

If you cannot explain why you DON'T use pooling for the variance
when Ho is true, then you really aren't much of a teacher of
statistics are you?

Here I am NOT saying you have an explanation and we disagree.
I am saying you DON'T have any explanation, let alone a valid one,
why you choose to ignore the Ho condition when the TEST STAT
is the distribution when Ho is true! See the two lines below:

You can beat around the bush all day long about all the other cases,
that would not add one iota to the p1-p2 = 0 case.

I continue to recommend that others read E&F and decide for themselves.

You not only beat around the bush, you are now EVADING the
bush altogether. I can't think o fa single statistics topic at any
level, let alone the low undergrad level that I cannot explain
without your kind of pointless excuse.


You even pulled the irrelevance of asking why the T test for
difference of MEANS don't pool the variance by combining the
two samples. There, you forgot that the T is based on the
mathematical statistics result of N(0,1)/sqrt(chi-square), two
independent variables of those two kinds. APPLES and
ORANGES you were misidentifying.

I raised it as a rhetorical question, you may recall. I knew the
answer. The reason I raised it is that the heuristic looks the same.

There is no heuristics involved with assume Ho true and pool the
variance
in the NARROW Ho p1 = p2 case.


Let me know when you've read the article. Otherwise, I've nothing more
to say except perhaps to repeat that I continue to recommend that
others read Eberhardt & Fligner (1977), The American Statistician, 31,
151-155, and decide for themselves.

You can repeat that till you die, Jerry, but I am NOT going to discuss
anything with anyone who has to rely someone else to read a useless
article and not be able to give his own reason for the SIMPLE problem
that is in every FIRST course textbook I've ever used.

That's not saying much about our educational system that produces
the kind of discussants in this newsgroup is all I can say to you,
Jerry.

That is probably your BEST tactical and nonstatistical excuse to
evade the issue. Good one! :-)

Others in this group, the NOISE makers, are likely to learn that new
way of making NOISE, demand that someone reads something
without being able to say what that something is. Even m00es
did better than you, Jerry. He quoted one paragraph from Hogg
and Craig that eventually turned out to be the paragraph that
distroyed his entire premise.

Adios, Jerry.

I'll be too busy to argue anyone the next couple of weeks anyway.

It took TOO much time to teach you that a MODE is NOT an average.
LOL It took me TOO much time to teach you that the computation of a
p-value REQUIRES one to know the Alternative Hypothesis, despite
that book by two misguided authors on the subject you kept quoting.

Time is too valuable to waste on someone who is supposed to be
educated, to teach them such elementary basic FACTS about
statistics that I am too impatient to repeat even for the "tree stumps"
undergrad in a Top 100 Ph.D. granting universities in the USA.

-- Reef Fish Bob.

.

Relevant Pages

  • Re: Testing the Equality of Two Population Proportions
    ... You haven't given one single reason WHY ... you don't use the pooled variance given all the reasons to use it. ... estimate the variance of p-hat when there is only ONE p-hat, ... you have not given one single valid reason WHY you want to ...
    (sci.stat.math)
  • Re: Stable Thread Timing?
    ... They just can just stop firing for no ... good reason. ... >> I have an application where I am using a System Thread to capture the ... >> debug.writeline) I am guessing up to an 80% variance on the time it takes ...
    (microsoft.public.dotnet.languages.vb)
  • Re: F-test for comparing variances?
    ... This usually comes up in an intro course in the context of a between groups t-test. ... The two group t-test is a special case of oneway analysis of variance, which is a special application of ordinary regression. ... We stick with the p/u/p represented by H0, until we have sufficient reason to believe otherwise. ... IMHO, it is unfortunate that terms like status quo, default, current, or presumed are not often enough associated with the "null" hypothesis. ...
    (sci.stat.edu)
  • Re: F-test for comparing variances?
    ... groups t-test. ... variance is by convention the larger of the two variances under ... Keeping H0 means that the [status quo, default, current, presumed] ... until we have sufficient reason to believe ...
    (sci.stat.math)
  • Re: fit F distribution
    ... The PDF function returned NaN or infinite values. ... your variable has mean 3.1338 and variance 4.5367. ... but didn't wind up with an F distribution. ... Is there a theoretical or other reason you need an F distribution? ...
    (comp.soft-sys.matlab)