Matti Narkia wrote:
> Thu, 05 Feb 2004 16:53:54 -0500 in article <
[email protected]> "Dr. Andrew B.
> Chung, MD/PhD" <
[email protected]> wrote:
>
> >Matti Narkia wrote:
> >
> >> Thu, 05 Feb 2004 13:50:16 -0500 in article <
[email protected]> "Dr. Andrew B.
> >> Chung, MD/PhD" <
[email protected]> wrote:
> >> >
> >> >Why are you interested in only elderly japanese men, Matti?
> >> >
> >> Is that rhetoric question? The referred study serves as an example that BMI probably should not
> >> be used as predictor of mortality in all subpopulations.
> >
> >Sounds like you are back-pedalling here.
> >
> Interesting. How did you get that impression? You must be _a lot_ more specific, if you want to be
> taken seriously.
You had set out to refute Nigel's assertion that:
"But BMI is still a quick and easy risk assessment for _all_cause_ mortality"
Now you are claiming that you had set out to refute that:
"BMI should be used as predictor of mortality in all subpopulations."
And, yet only you seem to be making this assertion.
>
> >>
> >> >Here are some better studies, including one looking at women.
> >> >
> >> Better? Different, but in what way better?
> >
> >(1) More people: 900,000 (1st study), 1,046,153 (2nd study), 115,195 (3rd study) versus 3741
> > (your citation).
> >
> As in food, more is not equivalent to better, especially if the purpose is to study a
> subpopulation, which may be smaller than the number cited above.
>
We are not discussing food here (yet
. Studies with more people have greater statistical power.
Some would see this as "better."
>
> >(2) More women.
> >
> So you like women?
Yes.
> Nothing wrong with that, so do I
.
Good recovery ;-)
(Mu was probably rearing up to comment about where Matt Bernstein had gone to
> But if the purpose is to research a specific male subpopulation more women in the study is hardly
> an advantage
>
However, for the purpose of generalizability of study results, more women (compared to no women)
is "better."
>
> >(3) Broader age range.
> >
> If the purpose is to study a subpopulation of elderly men, broader age range is hardly an
> advantage
.
Again, generalizability is an advantage.
> Your comments are getting somewhat hilarious
You seem to be the only one laughing
>
> >
> >> >
http://makeashorterlink.com/?X54D23F47
> >> >
> >> >
http://makeashorterlink.com/?I27D12F47
> >> >
> >> >
http://makeashorterlink.com/?U2AD42F47
> >> >
> >> These are large studies of general population, where deviating subpopulations largely cancel
> >> each others' effects.
> >
> >If that were true, there should be *zero* net effect.
> >
> Now here you have to be much more specific.
zero = 0 = null
> Please elaborate in detail, what you mean to expel my impression that you don't have a clue what
> you are writing about.
>
1-1=0
>
> >> They don't tell anything about validity of BMI in certain subpopulations.
> >
> >See the third study. Last I checked, women (or men) are a subpopulation.
> >
> Ah, so if the BMI seems to reasonably valid assessment of obesity and mortality in some
> subpopulations, the same must be true for all subpopulations? Chungish logic? Please elaborate.
Concept: Generalizability to an individual is more relevant than finding subgroups that are
seemingly exceptional.
>
> >
> >Weren't you trying to refute Nigel's assertion that:
> >
> >"But BMI is still a quick and easy risk assessment for _all_cause_ mortality"
> >
> I demonstrated that this cannot be generalized to all subpopulations, and that there are cases
> where there are better equally simple predictors.
>
Again, you are backpedalling.
Generalizability to the individual is more useful than finding subgroup(s) that are seemingly
exceptional.
Servant to the humblest person in the universe,
Andrew
--
Dr. Andrew B. Chung, MD/PhD
Board-Certified Cardiologist
http://www.heartmdphd.com/
--
Who is the humblest person in the universe?
http://makeashorterlink.com/?L21532147