Gus,

I  am glad you can do your thinking for yourself and do not need Paul to
jump in for you.

> Thank you for doing my thinking for me. Not!
> I did not "admit that CR works". I told you this in another message.
> I am still trying to figure out what is going on and have reserved
judgement.
> I do see that uniform samples exhibit the correlation patterns you
predicted.
> What to make of that, I do not yet know.


Ok. So CR works in that the polarization of the correlations between causes
does occur across the ranges of the dependent variable. You acknowledge
this. So you recognize that in the circumstance I define, that the dependent
variables can be distinguished from the independent variables.

>
> > Gottfried sees it as well, as have most others who bothered to do
> > the simulations. Gus's recent arguments have been based on an attempt to
> > create uniform distributions but his understanding failed to appreciate
why
> > uniform distributions were needed and why trimming was logical.  Gus
used
> > the word 'loaded" and that word is what prompted me to bring up the
issue of
> > honesty.  The word loaded usually means there is some hidden meaning
> > involved. There is nothing hidden in what I say.
>
> Yes, there is. You are twisting my words to fit your agenda.
>
I am not twisting you words. The data speak for themselves, at least when we
bother to explain how we get them. I have repeatedly asked you to show us
how you calculated your subsample and you are being evasive, pretending that
I am too dense to see what you did. Does you method fill out the corners of
the cross tabulation of x1 and x2? Are the intervals from which you sampled,
equal across the ranges of the variables?

Bill


>



.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to