RussellMc wrote:

>> I really disagree with the notion that correlation strongly suggests
>> causation. IMO this is one of the major pitfalls of dealing with
>> statistics. We /know/ (we =3D you + me + many others :) that there are
>> correlations that are not caused by a causal relationship, so a
>> correlation doesn't suggest anything -- unless there is something
>> else that shows that the correlation is in fact caused by a causal
>> relationship.
>=20
> Causation causes correlation.

Agreed.

> Correlation thus suggests in a given case that causation may be at
> work.=20

I agree. But "suggests ... that ... may be at work" is a long shot from
"strongly suggests", IMO.

It may also be that the correlation is due to both being (otherwise
unrelated) effects of the same cause. This is another thing that
correlation suggests, just as strongly as causation. In a complex system
where we only measure a very small part of the variables, it could be
that this is much more likely than causation.

> Increasing degrees of correlation increase the statistical probability
> that causation is at work.

IMO there's nothing inherent in correlation that makes a causal
relationship more probable than both being unrelated effects of a common
cause.


> If you get small clusters aka families, that by themselves are
> statistically too small to demonstrate a high probability of
> causation, enough such groups that either have high incidence or low
> incidence of a given disease, increase the chance that there is a
> causative agent at work. Or "if heart disease can be observed to "run
> in families" in a significant percentage of cases, then the existence
> of a causal element at the family level is suggested". eg - made up
> example, if 40% of the populace die of heart attacks then, if in 5%
> of families, 80%+ of people die of heart attacks, then there is a
> strong indication that there is a family level causative factor in
> these cases.=20

I follow all this.

> This does not HAVE to be genetic.=20

And of course I agree with this :)


> BUT, you make my point exactly. This study CLAIMS that in this context
> there are, apart from a few specific and well known exceptions,
> essentially NO things you can't change - THEY are, literally,
> claiming that a high % of family deaths aged in the  mid 70's due to
> heart attacks is NOT an indicator that other family members may be
> genetically predisposed to heart disease.

I sometimes ask myself where "died of natural causes" has gone. At some
point we all die, and when we do, it's because /something/ failed
fatally first. With medicine being what it is, we usually can pinpoint
some organ that failed -- if we spend the necessary effort. Considering
this, they may well have a point that a number of such heart attacks are
simply "natural cause" deaths that don't necessarily mean that there is
a significant family predisposition at all.=20

> IF this was true it would be stunning for me - it would mean that I
> should be able with due care and action to "probably" have a healthy
> outcome far different than my typical  family history - rather than
> just possibly.

"Due care and action" (or the lack thereof, in some situations :) is
probably the key...

Gerhard
--=20
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
.