In talk.politics.drugs Szasz <[EMAIL PROTECTED]> wrote:
> Brian Sandle <[EMAIL PROTECTED]> wrote in message
> <snip>
>>
>> I maintain training to be deficient.
> <snip>
> You know, I just wrote up several different posts detailing the
> laundry list of Sandle's logical fallacies, cheap rhetorical devices,
> and otherwise inane commentary, and then erased each one.
> I think it insults people's intelligence if I have to point out
> and detail how idiotic it is for you to try to lecture MD-PhDs and PhD
> candidates in clinical psychology in statistics, social sciences, and
> empirical design. So, please declare victory and go home, you're
> giving me a fucking headache.
You are still doing exactly the same, trying to claim quals as an excuse
for saying anything.
We should take the points that each person says and examine them, no
matter how lowly or high, with equal care.
It may be wrong that I have to draw attention to deficiencies. One of the
recent ones I detailed on talk.policy.drugs is the recent editorial of
BMJ, misreading as I feel they did (in Arseneault's report of effects of
cannabis youth use on later schizophreniform illness) `at age 15' as `by
age 15'.
The current deficiency I am looking at, that in statistical education, is
rather evidenced by what Erkki said as well as your `discourse'.
How can he say that Spearman's rank order correlation given by
rho = 1 - [6sigmaD^2]/[N(N^2-1)]
is just Pearson's product moment correlation given by
r =[NsigmaXY -(sigmaX)(sigmaY)]/SQRT{[NsigmaX^2-(sigmaX)^2][
NsigmaY^2-(sigmaY)^2]}
using ranks instead of scores?
An important difference is the distribution of data they are used on.
Going back to Bruning and Kintz it gives a couple of pages before the
method of Pearson product moment, the following:
Use [it] if your numbers represent amounts of some measurable
quantity--such as height,age, IQ, grade points, test scores, etc. This
analysis assumes that the two variables for which you have measures are
_linearly_ related. If you have ranked data, or are reasonably sure that
the data are related in some complex curved way, use one of the
correlation coefficients below. [My underline].
And those below are Rank-Order correlation, Point-Biserial correlation,
and Eta.
I do not think the book is making enough of the choice of which to use. At
least a couple of examples showing linear and non-linear relationships
would have been nice. Then maybe we would have avoided people like PhD
student szasz saying that Pearson's is the basis of everything in
correlation.
-----= Posted via Newsfeeds.Com, Uncensored Usenet News =-----
http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
-----== Over 80,000 Newsgroups - 16 Different Servers! =-----
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
