Re: A NEW publication: "Finnish Statistics for Dummies".


Anon. wrote:

What took you so long to post your NOISE this time?

Reef Fish wrote:
Anon. wrote: (18 mintes after RF's post):
Reef Fish wrote:
Anon. wrote (only 4 minutes after RF's post):
Reef Fish wrote:
Anon. wrote: (5 minutes after RF's post):
Reef Fish wrote:
<snip>

Yes, you're right. They're not. That's why the intercept still exists
even though it has a point estimate of zero.

Anon O'Hara forgot to mention -- ALWAYS, regardless of what
the data is -- that intercept will ALWAYS be zero.

That is a correct fact. But I can''t find no appropriate place to
put it other than in a new book "Finnish Statistics for Dummies"
since I had already a version "Portuguese Statistics for Dummies".

In Finnish Statistics, if you know there is something nonexistent
in a model (because its estimate will ALWAYS be ZERO no
matter what the data is), you can call that a "parameter" that is
known to be always zero.

Thus, if the TRUE model is Y = b X + e

Only if the empirical mean of Y is equal to the true (population) mean.
Are you really saying that this is always the case?

The answer to your question is the obvious NO, but your question
certainly qualify for the "Finnish Statistics" book.

But you'll have to wait for another installment, another edition,
or another chapter.

You're just dodging the question: in the model Y = a + b X + e, where X
is mean centred, how is the correlation between the estimated a and b
non-zero?

Baits for FREE tuition jettisoned into Finnish trash can.

-- Reef Fish Bob.

.

Relevant Pages