Re: ANOVA QUESTION - Terminology


Reef Fish wrote:
Richard Ulrich wrote:
Reef Fish Bob has offered several inappropriate comments
to this question, not to mention the insults aimed at me -

On 9 Sep 2006 17:28:52 -0700, "Reef Fish"
<Large_Nassau_Gr0uper@xxxxxxxxx> wrote:


Richard Ulrich wrote:
On 9 Sep 2006 12:53:19 -0700, "jp" <jpopovich@xxxxxxxxx> wrote:

I have the following set up:

One group (males, females)
One dependent variable (test scores), but tested pre and post tutoring.

Would this be a repeated measures design?

It can be analyzed that way, with the two periods.
RF >
The two periods are pre-tutor score and post-tutor scores and
it has only ONE variable, of the "difference" (improvement).

Depending on design, the "improvement" can be effectively
measured as the raw change score or difference; the regressed
change score (in the ANCOVA); or the simple outcome score
(rarely).





What would this be called (i.e., One-way ANOVA with repeated measures,
Two-way, etc.)?

For Pre-Post, the testing usually assumes that the group means do
not differ at Pre;
RF >
Why? Where did you get that?

That is sound advice which is offered in every course or text in
experimental design. Bob is apparently over his head here.

When Richard Ulrich mentioned something he said is "offered
in every course or lext in experimental design" and indelicately
offered that "Bob is apparently over his head here." that was
when I gave a brief summary of the graduate level courses in
Linear Models and Experimental I had taught.

And just how many of those courses have you TAKEN, not to have
known anything about Linear Models at the Neter et al level?

I have TAUGHT Graduate level courses in Linear Models from
books of Graybill and others, on experimental designs from several
graduate level textbooks such as one by Cuthbert Daniels, not
household name in statistics, but known by anyone who knows
much of anything about experiment design and and the analysis
of data based on those designs -- as follow-up courses to my
graduate level course in Data Analysis which has Neter et al's book
as prerequisite which Richard Ulrich had never reached the level.

In paragraph above, I mentioned I taught from a book by Cuthbert
Daniel, and characterized him as "not a household name in statistics,
but known by anyone who knows much of anything about experimental
design and the analysis of data based on those designs", I was
perhaps short-changing Daniel's recognition as a statistician, though
he was a self-taught one.

After my post, I went to the Google web and found this piece by
Stuart Hunter, on Cuthbert Daniel, that is much more deserving of
that man:

http://www.amstat.org/publications/tas/index.cfm?fuseaction=hunter1998

If the URL above gets truncated, try this one:
http://tinyurl.com/h98x5

I didn't realize that Cuthbert had died in 1997, and would have been
101 years old if he were alive today. His formal training was anything
BUT statistics, but his early associates had all the names of the Old
Guards that read like a Who's Who in statistics before 1950.

Quite a fascinating history of his education leading to statistics
also.

" Cuthbert's 1953 JASA paper "Statistics in Chemical Experimentation"
began as a review of a book. It proved so sweeping in its description
and condemnation of poor statistical practice that the editor, W. Allen

Wallis, published it as a contributed paper. Cuthbert left little doubt

that the industrial applications of statistics and the design of
experiments had come of age in both theory and practice. "

"Condemnation of poor statistical practice" has become a keyword
for the sci.stat.* groups since began reading in these groups in 2005.

Cuthbert Daniel condemned it in 1953.

Karl Perason condemned in much earlier, as I learned from one of
the participants in the sci.stat.math group.

In that respect, I find myself in good company when I play the
"watch dog" role in seeing the dozens and dozens of Quackery
committed by Richard Ulrich in these groups, and welcomed by
the innocent victims with open arms. Richard has done it again
in this thread.

Here's the condemnation long before 1950:

"As I grow older I feel more and more need not only for the censores
morum, but for censores scientiarum, a species of watch dogs of
science, whose duty it shall be not only to insist upon HONESTY and
LOGIC in scientific procedure, but who shall warn the public against
appearances of knowledge where we are as yet in a state of
ignorance. In this age of self-advertisement, when an individual may
become famous in twenty four hours by aid of the illustarated daily
press, there is QUACKERY in science as there is quackery in
medicine. And even where there is no quackery there is IGNORANCE and
DOGMA parading before the public as knowledge, and taking its TOLL
from the community by a multiplicity of devices. In many ways the
trained scientific man can WARN the public, even when it lacks
acquaintance with specialized detail... Unfortunately at the present
time no theory of what we may term scientific logic is taught to
students of science in our universities, and the result is only too
patent in 50% and more of so-called scientific publications."

(cited from E. S. Pearson 1938 Karl Pearson: An appreciation
of some aspects of his life and work. Biometrika, 29:161-248).

The poster, in sci.stat.math, who cited the Pearson passage said,

(no, it's not Reef Fish 2005 on sci.stat.math, and yes I took
liberty to capitalize a few random words :-)

My response was that I wish I could have said it half as eloquently
as Pearson did, nearly 70 years later, when I am seeing the same
Quackery and malpractice Pearson and Cuthbert Daniel saw in
1938 and 1953 respectively.

It's MUCH worse today because of the existence of Statistical
Packages and computing software!! It opened the door for
massive gatherings of Quacks who thought if they could make a
program run by stuffing some data in it, they have found a
solution to their statistical problem.

Need I say more?

I guess not. After the above addendum. :-)

-- Reef Fish Bob.


and the efficient test is usually the one-way
ANCOVA, using the Pre as covariate.
RF >
That's just a simple T-test for two independent groups on the
DIFFERENCE of the pre-and-post scores. Freshman/sophomore
stuff, Richard.

Simple change is *not* the same as regressed change.
Reef Fish Bob seems to have missed a course.

That is only because your own deficiency to realize that the
regression approach (if you knew HOW) to linear models and
experimental design accomplishes the same thing that are
much more clumsily done WITHOUT the use of Linear Models.


jp, get you an elementary testbook on testing means.

That may save you much time in unlearning some bad advice you
get from these newsgroups.

-- Reef Fish Bob.

-- Reef Fish Bob.

.

Relevant Pages

  • Re: Time To Teach Digital Etiquette
    ... they bully OTHERS by making incessant NOISE in the group. ... I spared naming Richard Ulrich this time because Richard ... Quackery, Malpractice, and Ignorance about Statistics, ... The other NOISE makers, m00es, Illywhacker, Bob ...
    (sci.stat.math)
  • Re: Scott Seidmans PURE NOISE on: "Non-linear regression problem"
    ... Richard Ulrich wrote: ... the substantive subject of nonlinear regression estimation is ... You failed to produce any evidence for your QUACKERY of using CV ... Statistics in 1938, long before I had used it, and he used it EXACTLY ...
    (sci.stat.math)
  • Re: Comparing two variation coefficients
    ... Richard Ulrich wrote: ... I thought about my usage there, taking "statistics" as ... "The test on variances of the log-transformed variable ... the equivalence in log-transformed variables ...
    (sci.stat.math)
  • Chief Quack of sci.stat.math resorts to his NAME SMEARING tactic of "Bob Ling" again
    ... statistics curriculm because if you know your statistics, ... Richard Ulrich and Bob O'Hara. ... insisting referring to me as "Bob Ling" even though everyone KNOWS ... who also had 9 hits when search for "Fuck you Bob". ...
    (sci.stat.math)
  • Re: Randomized block design
    ... > I use a randomized block design with 4 treatments and a control. ... > interaction between time and Ulva present/absent and the interaction ... Department of Mathematics and Statistics ...
    (sci.stat.math)
Loading