<HTML>ISHMAEL wrote:
[snip evidence of GSOH impairment and personal comments unrelated to the validity of statistics]
> Nevertheless, what you state here is not at all out of step
> with what I have said: our ability to measure always
> introduces a level of uncertainty.
But our ability to count does not. That alone gives the lie to your two ludicrously false statements:
> A theoretical *perfect* correlation (in the absolute sense)
> has a chance factor of zero
AND
> (but of course, *perfect*
> correlations never exist,
> A correlation can appear *perfect* when in fact, it is the
> incomplete nature of our measurements or samples which hide
> the aberations.
True, but that does not mean that all perfect correlations so suffer. Or, in logicians terms, if A implies B it does not mean that B is <i>only</i> implied by A.
> The "perfect correlation" in your example is
> achieved only due to the artificial limitations put around
> your sample.
Utter twaddle! It is due to counting. As indeed you do with sinks and divorces -- you <b>count</b> the bloody things, not <b>measure</b> them. Sheesh!
>
> -----------
> Correlation never implies causation (or relation beyond
> correlation);
> -----------
>
> You are correct in that correlation never implies causation,
> but you are incorrect in saying that it never implies
> relation.
But I didn't say that! I said <i>relation beyond correlation</>.
> In fact, correlation *always* implies relation -
> but only as a probability factor.
Twaddle. Calculating a correlation coefficient does not give you a probability.
> The closer the correlation,
> the greater the odds in favour of an existant relationship.
Again you are incorrect. A poor correlation may deny a relationship, but a good correlation does not support it <i>unless there is independent evidence for a functional relationship</i>.
Ishmael, if you are going to debate statistics, at least try to understand the rudiments first. The Moroney book that I recommended is really very good; it is also readable and inexpensive.</HTML>
