This subject is of great interest substantively and methodologically.

On genetic vs. environmental causes of Alzheimer's disease:
Africans living in France apparently do not show the pattern found
among African-Americans relative to Nigerians or native Africans.
Nor do Japanese living in Peru show the elevated Alzheimer's rates
found among those living in the US.  This may further pin point the
US environment as a contributory factor; or simply reflect the
higher reporting rates in the US.

Cavilli-Sforza reports that the genotypic variation among a small
population of Africans contains all the variation that exists among
humans.  A more diverse population will show some incidence of
almost any disease (or adaptive characteristic and talent as well).
Thus, the African-American experience tells us something about the
vulnerabilities of us all.

A methodological note.  I, for one, have a hard time with these
macronutrient measures.  They omit so much: vitamins, antioxidants,
glycemic index (not all carbs are equal), and so on.  Most high fat
eaters are also sweet carb browsers.  At a barbeque restaurant the
other day I ate ribs, a salad, and steamed vegetables, while most
everyone around me ate ribs, bread, and potatoes and sweetened beans
and drank cokes.  They ingested no fruit or vegetables and loaded
the glyco-oxidative pathway with sugar and fat.  It is a high
insulin response that triggers the eicosanoid pathway that produces
an inflammatory response (Protein Power by the Eades is a too-simple
source, but one I have at hand).

My colleague, the distinguished econometrician Jack Johnston
(Econometric Methods) is fond of saying that r square values can be
too high.  In a cross-country regression, anything above 0.7 should
be suspect.  Nature isn't that orderly and linear (untestable, I
know, but sound advice).  Values exceeding this level are often the
result of what he calls "partial identities" which are
transformations of the data that put the same variable on both sides
of the regression.  Something is fishy about these regressions
(speaking for myself and attributing no motives to anyone) and it
isn't prostaglandin 3.

But, what do they show anyway?  In the all-country sample, no
country is near the mean---it lies in that empty middle between two
far-flung clusters.  The regression and risk assessment express the
change in the expected value of the dependent variable (incidence of
Alzheimer's) conditional on a change in the mean of the independent
variable (fat or whatever).  The correct interpretation of the
coefficient is that if the fat intake among all 11 countries falls
by enough to lower the mean fat intake among all 11 countries by X,
then the mean incidence of Alzheimer's will fall by the coefficient
value times X.

Where this sort of calculation leads to bad policy is when it is
applied to a population or an individual.  There is no
representative population or individual to which this world sample
applies and it does not follow that your risks or my risks are
related in any way to our fat intake.  They may be, and the
mechanisms elucidated, along with the extensive cited literature,
may be a cause for reassessing one's own diet.

For all that, I think the Grant article is one of the most
stimulating I have read in some time.

Arthur De Vany
Professor
 <[log in to unmask]>
http://www.socsci.uci.edu/mbs/personnel/devany/devany.html
University of California
Institute for Mathematical Behavioral Sciences
3151 Social Science Plaza
Irvine, CA  92697-5100