--- [log in to unmask] wrote:
> Steve:
> >This leads to another observation. The Bayesian framework is, in my
> >opinion, preferred in that it is not dogmatic, allows for learning and
> >incorporates new information into the analysis. The PP just assumes
> that
> >the worst case scenario will occur and does nt allow for learning or
> new
> >information.
>
> It is always preferable to define what you are referring to. In this
> case
> talk about 'Baysian' statistics is merely talk. What you mean by
> Bayesian
> framework here is nothing new. Learning while doing means adaptation. Or
> adaptive management. There is nothing revolution, new and progressive
> about
> that.
In a sense your right. Bayesian statistics and decision theory have been
around longer than the current Frequentist techniques. However,
Frequentist techniques provide no mechanism for incorporating new
information into the analysis.
[snip]
> The pre-cautionary principle is a decision making principle. For
> example,
> the use of pesticides such as the phthalates for instance. In England
> there
> was a company that used (and may still) to make bug repellants with
> phthalates. Now I wrote a paper on the phthalates and I found there was
> a
> lot to be concerned about. The introduction of the substance into the
> diet
> of children is especially worrisome. In various test animals the main
> effect
> was testicular atrophy, etc.
The problem is, and which you have not addressed, is that what are the
risks that are involved. The Precautionary Principle (PP) assumes the
worst, that even a virtually undetectable amount of exposure is extremely
bad. This is why using a Bayesian approach is superior in my opinion.
You can look at the previous research formulate a prior and then as new
information comes available the prior is updated.
> If I was a planning on being a parent, I would not have any toys such as
> teethers, etc., near my baby. This is the application of the
> precautionary
> principle. To take precautions one should not knowingly allow the
> substance
> into the system of a fetus nor a baby. All toy manufacturers have
> removed
> this substance from their toys.
Well this does not prove much except that perhaps the toy manufacturers
are paranoid about lawsuits...actually a healthy paranoia if ever there
was one. Can you provide some information as to what level of exposure
these test animals had to recieve before they suffered from testicular
atrophy?
> The application of Bayesian statistics here has no purpose...not if you
> were
> to read my paper on the phthalates.
Uhhh, no that is wrong and a knee jerk reaction. A Bayesian Decision
framework would allow you to include previous research and information on
phthalates. Then as new data is available the results are updated. If
you want to err on the side of caution you can do so by selecting a prior
distribution that accounts for this and then let future data inform you as
to whether or not this initial fear is reasonable. By choosing the PP you
have absolutely no possibility to learn or gain new information. All you
are left with is your initial position which is as Jim noted, extreme.
Steve
__________________________________________________
Do You Yahoo!?
Send instant messages & get email alerts with Yahoo! Messenger.
http://im.yahoo.com/
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|