NetNews Usenet Archive 1992 #31/NN_1992_31.iso/spool/sci/med/22947

Text File | 1992-12-22 | 6.6 KB | 133 lines

Newsgroups: sci.med
Path: sparky!uunet!cs.utexas.edu!sun-barr!ames!purdue!mentor.cc.purdue.edu!pop.stat.purdue.edu!hrubin
From: [e-mail] (Herman Rubin)
Subject: Re: Bashing, truth, etc.
Message-ID: <[e-mail]>
Sender: [e-mail] (USENET News)
Organization: Purdue University Statistics Department
References: <[e-mail]>
Date: Tue, 22 Dec 1992 15:56:30 GMT
Lines: 121

In article <[e-mail]> [e-mail] (David Nye) writes:
>[reply to [e-mail]]

>>I am afraid you misunderstand both what I said and also what is
>>involved in "scientific" medicine.  I am putting on my professional hat
>>here.

>I will defer to your expertise in the area of statistics and its
>appropriate application to scientific studies, but your statement:

>>I do not need any data whatever to know that saccharin has an effect on
>>cancer rates, or that vitamin C has an effect on longevity, or even that
>>the most disreputable quack treatments have an effect.

>is not exactly what the rest of us would consider consistent with
>scientific medicine.  If these statements are true, it should be
>possible to design studies to prove them true.  If the beneficial effect
>of some treatment is unmeasurable, then its utility is negligible.

Read the statements more carefully.  I said that I do not need any data
to show that it has an effect.  This means that the effect is not
EXACTLY zero.  This statement says nothing about the magnitude of
the effect.

The Canadian rat study on saccharin is a good example of the type of
problem involved.  If one gives a large enough dose to enough subjects,
a "statistically significant" difference will be found.  If one even
gives a small dose to enough subjects, likewise.  The number of subjects
needed to have a good chance to find statistical significance may be
quite large, but this has absolutely nothing to do with the magnitude
of the effect.

>>The problem is how much is the effect, is the effect different for
                 ^^^ ^^^  ^^ ^^^ ^^^^^^
>>different individuals, how does it interact with other things, etc.
>>The problem is not an easy one which can be handled with a simple
>>statistical package.

>Failure to come up with an appropriate statistical method to analyze the
>results of the study is not a failure of statistics or of the scientific
>method but of the study's design.  I suspect you have too often been
>asked to come up with a plan for statistical analysis after a study has
>been completed rather than at the more appropriate time when the study
>is being designed.  In the case where a treatment has a measurable
>effect only on a few subjects, it would be reasonable to identify
>responders in an open-label trial, then restudy that subpopulation with
>a double-blind, placebo-controlled crossover design.

The failure is more of a misunderstanding of what can and should be
inferred from data.  That point null hypothesis is ALWAYS false, and
I do not need any data for this.  The problem is not about rejecting
a true point hypothesis, but about which decision to make; should one
act as if the point hypothesis is true, knowing that it must be false,
and having a simpler model available, or not.  Given enough data, any
hypothesis will almost certainly be rejected.  But without acting on these
necessarily false simplifications, all one has is the data, and there is
no reasonable method of going from a miscellaneous collection of data to
a reasonable course of action.  The classical statistical procedures of
alpha-level testing, P-values, confidence intevals, etc., are not well-
founded and cannot be.

But a decision-theoretic based approach always works in principle.  It
may have problems of complexity and computation in practice, and these
problems cannot be ignored.  You place too much faith in the ability of
classical statistical procedures to come up with an impossible "truth".

>>A professor of nutrition may make materials available to me, but is not
>>allowed to give advice by law.  There are many actions which are
>>illegal, even if all the cards are laid on the table.  Most of them
>>should not be so.

>In other words, you are recommending that the practice of medicine be
>deregulated?

I am recommending that the use of credentialism as a substitute for
competency be eliminated.  A physician whose knowledge of nutrition is
essentially zilch can give nutritional advice, and many of them do as
a matter of course.  In the nutrition group, there was a statement from
a professor of nutrition in a medical school, himself an MD, comparing
the nutritional knowledge of other MDs who are faculty in that medical
school with secretaries.  His opinion was that there was not much
difference, with the medical faculty having a slight advantage, unless
the secretary has been on a diet, in which case the advantage would go
the other way.

I have been a professor of statistics for many years.  I do not believe
in the use of course grades or credits or degrees as more than, at best,
an indication, and not a good one at that.  I have never myself had a
statistics course of any kind.  This does not affect my status in the
academic and statistical communities.  Those who know should be allowed
to share their knowledge, as long as they do not make false claims.  Those
who do not know should not claim to be adepts.

>>In addition, there is effectively a conspiracy by government regulators
>>to make information not available.  I have no way of finding out the
>>information about a drug which I would consider appropriate for an
>>intelligent decision, nor does a physician have access to that
>>information.

>What information are you speaking of?

The detailed information available as to the effects of the drug, including
numbers of cases.  With this information, I can look at the probabilities
as a function of the state of nature; without them, I cannot.  Nor can
anyone else; this is known in statistics as the likelihood principle.

>>Prepaid medical care, whatever it is called, is a socialist conspiracy.

>Ah, now I think I see where you're coming from.  You're a libertarian,
>no?  I recognize the refrain.  I'm a liberal democrat.  We are at the
>same end of the libertarian-totalitarian axis

I do not see how you can say this.  No libertarian, or even a classical
liberal (what was called liberal before the socialists preverted the
term in THIS century) can consider democracy as other than totalitarian.
But this is for the political groups.
--
Herman Rubin, Dept. of Statistics, Purdue Univ., West Lafayette IN47907-1399
Phone: [phone]
[e-mail] (Internet, bitnet)
{purdue,pur-ee}!snap.stat!hrubin(UUCP)