On Fri, 29 Dec 2000 22:42:18 GMT, [EMAIL PROTECTED] (Alf
Breull) wrote:
> On Tue, 26 Dec 2000 16:32:54 -0600, jim clark <[EMAIL PROTECTED]>
> wrote:
jim > >
> >There is a considerable literature on clinical judgment (i.e.,
> >interview and human judgement) vs. actuarial predictions (i.e.,
> >predictions from demonstrably valid regression equations ...
> >human judgment _might_ be used in producing individual predictor
> >scores, but not in aggregating them). In general, human judgment
> >does not fare all that well relative to actuarial (i.e.,
> >statistical) methods. Interesting that someone posting to a
> >statistical newsgroup would advocate the non-statistical approach
> >to selection problems.
>
Alf >
> This is one possible view, depending on your psychological education,
> background and experience.
Alf,
To me, that seems like a damnably condescending statement, from you.
Jim's statement is not a "view"; it is 100% correct: there *is* a
literature, and that does not depend at all on "your psychological ...
background." What makes your part "damnable" is that Jim wrote
neatly and intelligibly, whereas your note has parts that are dubious
and other parts that I can't even parse.
> Coming from a methodological background, you could, without problems,
> offer lots of regression equations which were errorously claimed as
> 'demostrably valid' one or other day, scientific publication or
> period.
- for instance, what is <the above> supposed to mean?
- that some idiot could offer pseudo stat- reasons for something, and
that is exactly how Jim's note looks to you?
- If you meant that, please read better; if not, please write better.
> Coming from test psychology, you would hesitate because reliability
> and validity coefficents, usually, suffer from 'regression to the
> middle', are too low for individual prognosis, explanantion (in the
"Regression to the middle" is not a complaint I have heard assigned to
reliability/validity coefficients, in five years of reading 4 online
stat groups. The single largest problem: the coefficients describe
both the sample and the scale, and "norming samples" may be
ill-suited to the uses people make of them.
< snip, rest >
I think you were trying to make some cute point, concerning the
over-reach of scholars, but -- it is my impression --
a) you don't know the fields well enough, and
b) once you know them, you would not have the point to make.
c) Actually, I am concerned with over-reach, too; but I think the
target must be in a different direction.
It might be that some single, honored elder believes an exaggerated,
overstated view about "brain lateralization" and gets quoted in the
newspaper. Or tabloid. Most of the extreme quotes, however, will be
from non-experts, or will be resented by experts who were misquoted
(unless, maybe, it serves the purpose of attracting funding for more
research.)
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
