Re: Does Reefer know any math beyond elementary calculus?


illywhacker wrote some nonsense just to cover up the track of
his HISTORY of posting errors about Bayesian Statistics.

Reef Fish wrote:

Below are Illywhacker's posts, NUMBERED by Google in the thread

1. The OP, AB's post was on Fri, Oct 6 2006 6:47 am
2. My initial reply (I cited) was on Fri, Oct 6 2006 11:00 pm
3. Your mention of MAP (I cited) Sat, Oct 7 2006 7:32 am
This was the BEGINNING of Illywhecker's NOISE about
what I posted.
66. "See my reply to Reefer" Sat, Oct 7 2006 7:33 am
5. Illywhacker Sat, Oct 7 2006 8:08 am
7. Illywhacker Sun, Oct 8 2006 5:17 am
10 Illywhacker Mon, Oct 9 2006 8:45 am
12 Illywhacker Mon, Oct 9 2006 12:00 pm

That was the FIRST time Illywhacker even responded to the OP,
about some irrelevant programs in R. AB ignored him ever since.

To re-iterate, you FALSELY accused me of introducing MAP. You
had your facts and chronology all wrong, and my documentation
about your posting HISTORY on the Bayesian inference topic is
exactly correct, as I had posted, repeated for convenience:

In your response to Lou Thraki, you wrote,

LIE> 1) The MAP estimation issue came up as RF's interpretation
LIE> of the OP's question. I am neither supporting nor opposing it.

Read what was EXACTLY said in my initial response to AB, in

http://groups.google.com/group/sci.stat.math/msg/3631d117792e2c36

Then Illywhacker posted paragraphs of gibberish about
differentials and derivatives that had NOTHING to do with
Bayesian priors about inference about the mu and sigma
of the Normal distribution.

I have not seen L.J.Savage nor any Bayesian talk about ANY of
Illywacker's hocus pocus use of derivative and differentials in the
problem of Bayesian inference about the parameters of a normal
distribution. That includes all the textbooks from which I've
TAUGHT Bayesian inference about the Normal parameters mu
and sigma^2, the latter is often reparametrized into it's reciprocal
to suit the conjugate prior. That was about it. None of
Illywhacker's WHACKING with his misguided mathematics
practiced by physicists, in the name of Bayesian inference.

Illywhacker LIED about me bring up MAP, when he was the one
who did it (proven above), and he had been talking about MAP,
MAP, MAP, while he WHACK, WHACK, WHACK, on something
that is completely foreign to Bayesian inference by Bayesians!

LIE> So, as RF says, if you want to MAP estimate, and if you use an
LIE> ignorance prior, MAP estimation, done correctly, reduces to MLE,
LIE> which seems very reasonable.

The above is your ultimate LIE and misrepresentation.

I NEVER said nor implied ANY of what you said in those three lines.
I NEVER even used the term MAP in which you keep digging in
your hole of ignorance. I merely mentioned the FACT that a
"diffuse prior" (apparently a term Illywhacker has never heard
before that has very definite and technical meaning in Bayesian
inference), mentioned numerous times in the writing of Jimmie
Savage, Dennis Lindley, and the rest of the well-known Bayesians
on both sides of the Atlantic. A diffuse and uninformative prior
means there is NO integration of ANY prior into the likelihood
function, and hence the posterior distribution IS proportional
to the likelihood function itself, and therefore the MAXIMUM of
such posterior distributions is precisely the MLE of the non-Bayesians.


It's all standard facts and knowledge about Bayesian Inference
in Bayesian Statistics. ALL of which is foreign to:


It ALL started with Illywhacker's introduction of MAP, completely
unrelated to anything I talked about in my posts, starting with


http://groups.google.com/group/sci.stat.math/msg/3631d117792e2c36

I asked AB what he meant by HPD because this was his entire post
was asking about HPD, and nothing about Illywhacker's MAP,
and the SUBJECT posted by AB was "Highest Posterior Density".

The rest was history.

Illywhacker brought in his favorite toy MAP (and apparently the
only toy he knew).

As I said, I didn't even know what Illywhacker's MAP was.
AB's HPD is not a well-known acronym either. If you look for HPD,

you'll find Houston Police Department, Household Product Database,
Houston Police Debate, and lots of other topics -- I never got to
the Bayesian part.

But HPD is used in conjunction with Bayesian Inference. See, e.g.

http://citeseer.ist.psu.edu/chen98monte.html

It's about INTERVAL estimation, not POINT estimation, as in MLE.

It's about Bayesian Credible Intervals (a term todistinguish from
the non-Bayesian Confidence Interval).

So, Illywhacker not only doesn't know that HPD is about
INTERVAL estimation, uses the term CI instead of Credible Interval,
and kept talking about MAP, which is a statistical term used in
conjunction with tomography reconstruction in IMAGE
processing:

http://citeseer.ist.psu.edu/context/92838/0

That probably relates to some of his work in PHYSICS, but that's
about all.

So, here we have a complete HISTORY of Illywhacker's
lack of understanding about Bayesian inference in Baysian Statistics,

from HIS choice of

MAP which is a WRONG characterization of HPD
(see web references above).

MAP is an image processing use in tomography somehow got the
name Bayesian attached. It is certainly NOT associated in any
way with Bayesian Credible Intervals nor Highest Density
INTERVALS in the posterior distribution.

From, that point on, Illywhacker merely dug deeper and deeper
into his mathematical hole, bringing in the irrelvance of
Riemannian metric, Lebesque integration, and his use of
differentials in UNINFORMATIVE priors in multivariate
settings that Bayesians don't know how to represent
informative priors in high dimensions.

This pretty much summarized, as well as EXPLAINED
the source of, Illywhacker's misguided notions about
Bayesian inference, their use of priors, and their use
of posterior CREDIBLE INTERVALS around the
highest point of the posterior density, which certainly
makes much more sense than just finding the maximum
POINT, the analogue of MLE.

Illywhacker try to cover up his posting errors and false
accusations by more and more derogatory SUBJECT like

"Does Reefer know any math beyond elementary calculus?"

or his previous one,

"Reefer fails to read posts again "

or other earlier ones,

I'll simply RE-POST this entire post in another new thread,
if necessary.

-- Reef Fish Bob.

.

Relevant Pages