I blundered twice (the second time in thinking I'd blundered the first time which I suppose means I only blundered once.) The present example is not a binomial as we are concerned not with the number of sales at the drive-in window but the percentage of sales dollars. To apply permutation methods, we need to compute the 50 daily percentages, subtract 75% from each, and then test the null hypothesis that the resultant distribution is centered and symmetric at 0. [The need to assume symmetry would make me reluctant to apply permutation methods in this instance.] The point is that we can test non-null primary hypotheses directly via other methods [though come to think of it if you want to treat those daily percentages as if they were normally distrbuted, you would have to assume they were symmetrically distrbuted also.]
Phillip Good
Donald Burrill <[EMAIL PROTECTED]> wrote:
Donald Burrill <[EMAIL PROTECTED]> wrote:
(Reply to OP and to edstat.)
On Tue, 24 Feb 2004, Phillip Good wrote:
> Null means null.
One can hardly argue with that. Now, let me point out that one of the
meanings of "null" is "zero". Are we to understand, Phillip, that you
are arguing that the proper null hypothesis in the case under discussion
is H0: P = 0 ?
If this is what you mean, please show us how you would test it. I want
to watch.... But perhaps this is not what you had in mind. (Hard to
tell -- telepathy is not one of my skills.)
> A terifying habit is to state in error that the primary hypothesis is
> null when it is not.
Are we also to assume that "terifying" means "terifying" and that this
is not the same as "terrifying"? Then a definition would help. But
perhaps it's just a misspelling...
> For example:
Example of a habit (terrifying or otherwise, as may be)?
Example of a state[ment] in error?
Example of a primary hypothesis that is not null?
None of these seem to describe what I perceive as the situation
presented by "burt" (<[EMAIL PROTECTED]>) and addressed by various of
our colleagues including Dennis.
> Dennis Roberts <[EMAIL PROTECTED]>wrote:
>
> "I would say the null is .75 ..."
>
> What he meant was that the primary hypothesis was that drive in
> percent =0.75.
Well, no, and notwithstanding his later recantation (if that's what it
was ;-), Dennis was [correctly, IMO, if somewhat telegraphically]
asserting that the null hypothesis for this situation is that what you
chose to call "drive in percent" (which puzzled me for a moment, till I
figured out that you didn't really mean "Drive, in %", but rather were
using a kind of shorthand for "drive-in revenue as % of total revenue")
has the value .75. And you have mis! -stated the value by a factor of
100; the null-hypothetical percent would be 75, not 0.75. (Agreed,
Dennis wrote ".75"; but he didn't write "per cent" with it.)
> This distinction is essential when we come to choose among methods of
> testing. Hypotheses must always be converted to null form before
> permutation methods are employed.
Are we choosing among methods? I didn't notice but one method having
been mentioned, but perhaps I wasn't paying attention... Or perhaps you
are asserting in effect that one is necessarily choosing among methods,
although the choice may not be consciously made?
> One also ought take in consideration the alternate hypotheses before
> making a recommendation as to the method and statistic to employ.
> Are they one-sided or two sided? Ordered or unordered? Is the loss
> function first-order, second-order, mini-max, or something else?
These sound like terribly arcane and recondi! te questions to be
addressing to poor "burt"; and it doesn't appear to me that any of the
rest of us have adequate information to frame reasonable answers. But
perhaps you were attempting to address the underlying theory of pedagogy
associated with statistical consultation, and leaving it to the rest of
us to figure out how to translate these concerns to "such a Tongue as
the people understandeth": the people in this case being "burt" at
least, and possibly also his client(s)m unless the problem was (as it
may have been) a homework problem perpetrated by another of our
colleagues.
> Phillip Good
>
> author, Common Errors in Statistics and How to Avoid Them.
> "Never trust anything that can think for itself if you can't see where
> it keeps its brain." JKR
This too is interesting advice, and leads more or less naturally to the
question, what such thinker do you know for which you CAN see where it
keep! s its brain?
Ciao! -- Don.
------------------------------------------------------------
Donald F. Burrill [EMAIL PROTECTED]
56 Sebbins Pond Drive, Bedford, NH 03110 (603) 626-0816
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
Do you Yahoo!?
Yahoo! Mail SpamGuard - Read only the mail you want.
