Re: Richard Ultich the sci.stat.math resident Quack is beating his Dead Horses AGAIN


Richard Ulrich wrote:

See the "subject"!!

As Yogi would say, "It's deja vu all over again".


> > > > Note that Ulrich NEVER cited any specifics, only HIS verbal
> > > > version of distortion -- which is Ulrich's modus operandi, in his
> > > > attempt to defend his Quackery and Malpractice in statistics.
> >
> > > > < Richard Ulrich's self-serving distortions snipped -- interested
> > > > readers can, and SHOULD, read those threads themselves!!
> > > > Without any coaching from Richard, or ME. >
> >
> > That should have been the END of it.


Now regarding the actual question by the OP and Richard's FINAL
follow-up to the OP, with hte benefit of a roadmap drawn by me.
> >
> > After I had already given the OP the roadmap about where he went
> > wrong and how he should have proceeded, Richard Ulrich then
> > followed my script TWO DAYS LATER (though not quite successfully,
> > because Richard obviously didn't understand it himself) to give the
> > sehwail further plagiarized advice, in which Ulrich's muddled mind
> > showed!
> >
> > RU> [snip, most of the stuff about normality testing, etc.
> > RU> These are not essential to regression or to the Question
> > RU> being asked, so long as you have confirmed that all the
> > RU> data values are legitimate and reasonable, with no big
> > RU> outliers to confuse whatever is going on.]
> >
> > "not essential"? Absolutely unnecessary and irrelevant!
> >
> > I had already said it plainly and unambiguously to the OP two days
> > before Richard's attempted (but misplayed) plagiarism:
> >
> >
> > OP> A. Normality
> > OP> I started with checking the normality of the IVs and DV
> >
> > RF> WHY? There is absolutely NOTHING in any
> > multiple-linear-regression
> > RF> model that requires ANY of the Ind. vars. or the Dep. Var. (which
> > is an
> > RF> aggregate of n observations from n DIFFERENT distributions, even
> > if
> > RF> you assume each of them is normal) to be Normal.
> > RF>
> > RF> Therefore, you work below is precisely Garbage In, Garbage Out,
> > RF> because all the numbers are 100% irrelevant to any regression
> > model
> > RF> you are trying to fit. It's a complete waste of computer and
> > human time.
> >
> > Let me emphasize the "100% irrelevant" to any model and any problem
> > of the type the OP was attempting.
> >
> > After bungling my advice about the absolute lack of necessity om doing
> > any
> > plot looking for normality, Richard Ulrich improvised this part
> > himself:
> >
> > RU> with no big outliers to confuse whatever is going on.]
> >
> > This is an unmistakable manufestation of Ulrich's OWN confusion about
> > those plots sehwail did, and my advice that they were 100% useless.
> >
> > Outlier in WHAT? Any outlier in ANY of the independent OR the
> > dependent
> > variable is 100% irrelevant to the probability model about the ERRORS
> > (residuals) of the model, because no model had been fitted at that
> > stage!
> >
> > That's the kind of Quakery characteristic of Ulrich's comment on any
> > linear model or regression problem. It sounds learned and reasonable
> > in his muddled mind, but absolute GARBAGE to anyone in the know.
> >
> ======= back to commentary --
> > I read Richard's bungled advice soon after he posted it, but decided
> > to let it go. Until Richard Ulrich felt compelled to bring out some
> > DEAD HORSESl that were entirely irrelevant to the PRESENT
> > THREAD or the PRESENT PROBLEM, and dwelled on his own
> > distorted fantasy on them, rather than addressing the actual problem
> > at hand or let the readers find his alleged fault of mine in his Dead
> > Horses.
> >
> > Richard Ulrich's FINAL critique, comment, and advice (after having seen
> > mine) would have gotten him a grade of no better than 30 out of 100 if
> > he were in my data analysis course, for not knowing, and failing to
> > consider many of the RELEVANT steps in exploring nonlinear surfaces,

>
> However, here we can agree on a couple of things. For one,
> there are posts that don't need comment, because they do no
> harm. For another, my "critique" surely would be inadequate
> for an answer on response-surface fitting in a data analysis course,
> BECAUSE that is not what I was doing. It is rather unlikely, in
> my opinion, that this poster wanted or needed to do non-linear
> fitting. It became evident, in my probing, that he isn't reading
> his regression output correctly, and when he mentioned "nonlinear",
> it was probably just grasping at straws.

That was only ONE of the issues in model building!

You conveniently neglected to defend your ERRORS in advising
him to look at normality and outliers in the scatterplots of the
INDEPENDENT varuabkes and the DEPENDENT variable, when
he was looking for a multivariable regression fit, LINEAR OR NOT.

That was the absolutely unnecessary part that you failed ot
understand, as the OP did not know but probably understood
NOW, since by his later account, he certainly seemed to have
read more regression books and journal articles than Richard
Ulrich ever had.

>
> Bob wrote a fine reply, technically. I said that it might or
> might not be useful [to the OP] -- My notion was, that
> answer was far above his head, and not at all what he needed
> to know right now.

Perhaps. But he certainly does not need Richard Ulrich to make
the same blunder he did as in the "normaility" and "outlier" issues,
not to mention the serious omission!

Isn't it convenient to gloss over those errors of yours, Richard?


and this is all Richard could come up this round:

> - Similarly, I condemn Bob Ling for his absolute
> lack of empathy and understanding in giving
> statistical consultation to a struggling beginner.

This comment was actually Richard Ulrich's self-recommendation
as a consultant, because he has lost his job and is looking for one
NOW.

I can give anyone looking for a consultant for STATISTICAL
advice on theory, methodology, and practice that hiring
Richard Ulrich for such a job would be like hiring a drowning
victim in a shallow pool to be a Life Guard in a deep ocean!

I mean it SERIOUSLY, based on the numerous posts by
Richard Ulrich I've read and discussed during a period of
about 6 month's of participation in rec.stat.math/edu groups.

The evidence was AMPLE and Unequivocal!


Everyone (even Richard Ulrich) could see that sehweil was a
"struggling beginner", and I advised him accordingly. after giving
him a LENGTHY explanation of his (sehweil's) errors to sehweil,
in my initial response:

RF> In your case, I strongly advice taking an elementary course in
applied
RF> data analysis in model building, and not try to do-it-yourself
after
RF> reading a few too-simplistic articles or posts in newsgroups.

But Richard Ulrich failed to recognize his OWN "struggling beginner"
status on the subject, and tried his "blind leading the blind trick"
to which Richard has grown accustomed.


> > Anyone who is half way competent in data analysis and model building
> > could read Richard Ulrich's FINAL analysis for sehwail, and see for
> > themselves why Richard Ulrich is an incompetent Quack on the
> > subject.
>
> Let's see -- Bob Ling does not have the empathy to be
> much of an educator, or, especially, to be a consultant
> to people who are asking the wrong question.

People ask the wrong question(s) all the time. Educator or
consultant, there is NO EXCUSE for compounding his wrong act
of trying to look for normality in the scatterplots by adding
YOUR own wrong acts/advice of looking for outliers and that
"normality is not essential" when it was absolutely UNNECESSARY!
>
> I can leave the other Sleeping dogs lie, for now.
>
> ====== no comments on the rest.

My comment to Richard Ulrich stands, and clarified and emplified
in this post, relative to sehweil's problem in the sehweil thread.

-- Bob.
> >
> >
> > Richard Ulrich, give it up.
> >
> > You are bringing in your distorted Dead Horses only as your way of
> > trying to hide your inadequacy and malpractice in sehweil's problem.
> > OR, it is actually quite possible that you don't even KNOW how
> > ridiculously erroneous your adive to sehweil was! Or WHY!
> >
> > I had told you many times before: your time will be better spent
> > LEARNING statistics and multiple regression yourself FIRST, with
> > the hours of tedious excuses posted for your own rationalization
> > and excuse.
> >
> > You should pay need to the First Law of Holes,
> >
> > "When you find yourself in a hole, stop digging."
> >
> > That's my advice to you, RIchard Ulrich, the sci.stat.math resident
> > Quack
> > of statistics.
> >
> > -- Bob.
> >
> > Taking a few minutes from the cruise away from Montego Bay,
> > Jamaica, as my public service to combat statistical Quackery and
> > malpractice, exemplified by Mr. Richard Ulrich, formerly
> > Assistant Professor of Psychiatry, last verified (a few months ago)
> > as no longer employed in said department.
> >
> > The public service would not have been taken had Richard Ulrich
> > not chosen to post his two lengthy, distorted, repeat of his Dead
> > Horses entirely false and faulty as well as irrelvant to the present
> > thread and discussion!
>
> --
> Rich Ulrich, wpilib@xxxxxxxx
> http://www.pitt.edu/~wpilib/index.html

.

Relevant Pages