In article
<9ce0e184-18ba-48ef-839f-60bf5754eb9f@[EMAIL PROTECTED]
>,
Hannele.Tervola@[EMAIL PROTECTED]
<Hannele.Tervola@[EMAIL PROTECTED]
> wrote:
>On May 6, 10:54 pm, hru...@[EMAIL PROTECTED]
(Herman Rubin) wrote:
>> There is not just the problem of observations, but the
>> inferences made from the observations. It is NOT the
>> case that the observations determine the inferences made.
>If you are higly skilled in objective thinking, you can make the basic
>deductions right and so predict what even the UFOs must have known in
>order to build the kind of space****ps that they do - IF they exist...
>> In simple cases, it may appear that this is so, but when
>> the cases are not so simple, good scientists can come up
>> with widely divergent models. It is at this point that
>> one can show that objectivity is not possible.
One can construct VERY SIMPLE models so that NO objective
approach can come close to matching the performance of
non-objective approaches.
>Objectivity IS possible, different people are just good at different
>sides of it, maybe entirely bypassing the points that others rise.
The "best" type of inference is to produce a family of
mathematical models, with parameters in them, and to then
analyze the fit, the value, and the results. The above
problem occurs in reasonable simple models.
>When I was on a course in theoretical biology, there were two types of
>models: complex ones including all the phenomena that were known to
>affect things and simple ones with only the factors big enough to be
>measured counted. The former brought much more understanding, the
>latter were easier to verify by scientifical field experiments (they
>were like descriptions of the data from field experiments) rather than
>by common sense and long chains of deduction of what phenomena ought
>to be there (i.e. deductions about what the world is like when
>estimated in the light of ALL of our knowledge). Both were correct,
>but the unfortunate thing is that they did not understand each other:
>they were educated persons rather than thinking ones, so they failed
>to notice what was not said explicitly: the different good sides of
>these two methods, they suitedness to different things.
You are giving a major example of non-objectivity. An
objective approach would unify the two models. It would
then completely change the theoretical biology course so
that one would not go experiment -> theory, but theory
confirmed by experiment.
The announced theories are all there, just waiting to be
dragged out. But they have different degrees of obviousness,
and thus are not equally likely to be produced. So that when
the data call for a "complicated" theory, as is often the case,
one has to trot out the theories first, and see which give a
reasonable fit. Whatever theory is announced is WRONG, and
there are situations where a theory which gives a worse fit
is preferable to one with a better fit. Not only that, two
intelligent investigators given all the information and all
the analyses might make different choices, and both are doing
the right thing. There goes your objectivity.
>> This lack of objectivity is treated in statistical decision
>> theory, and the answers are often that the answers are not
>> easy. This is not learned by starting out with apparently
>> simple cases, but by "plunging into the deep end."
>In my opinion one should first figure out what there is in the
>reality, what can be understood about it, rather than to claim that
>nothing can be understood, so let's start with the obscure cases where
>there is no way to verify your results against the real dynamics of
>the world. Statistic just describes the average, the fluctuations and
>the propability distributions of the individual occasions.
You clearly have no understanding of statistical theory,
and only know of techniques which have MANY restrictions
to be valid. In fact, some of the heavily used techniques
are rarely of any validity at all. Learning formulas does
not help at all.
Statistics
>does not tell anything about the dynamics of the world, unless of
>course that is born out of having great m***** of something, like
>atoms or people for example, so one ought to start with simple
>situations from which one can find the dynamics and then with that
>knowledge count the statistical estimates of what will happen in
>complex situations with a great number of factors affecting.
See my preceding paragraph. Learn statistical theory,
and forget, or at least put in the far-distant corner,
the methods you have been given as dogma.
>> Of course, one will need flotation devices and diving bells,
>> and so the abstract concepts of logic, mathematics, probability,
>> and "mathematical" statistics are needed. It is the concepts,
>> not the ability to solve simple problems, which are im****tant.
>> Knowing what addition means and when to use it is im****tant;
>> knowing how to perform the addition is useful, but not very
>> im****tant.
>I have a very different view: What matters in thinking, is the forming
>of a correct holistic picture of the world. Concepts are just tools
>that we use to pay attention to certain charachteristics of the world.
>In this I agree: it is good to know what kind of phenomena there are
>in the world, for that concepts are good, but the final aim of
>thinking, the only thing that matters is the forming of a good
>holistic picture of the world. And that too is relavant only from the
>point of view of being able to guide our actions, so putting things to
>practise is very im****tant indeed!
Precise thinking is mathematical for problems in which "truth"
can be found. It is statistical if there is some form of
uncertainty; that is what statistics is, not the methods of
dealing with ultrasimple special cases.
Both mathematics and statistics cover more than you understand
of them. As many have stated, if it is precise, it is mathematical.
--
This address is for information only. I do not claim that these views
are those of the Statistics Department or of Purdue University.
Herman Rubin, Department of Statistics, Purdue University
hrubin@[EMAIL PROTECTED]
Phone: (765)494-6054 FAX: (765)494-0558
|
