Religion VS Science
Axel of the North!
After seeing the Daily Show's weeklong "Evolution Schmevolution" and
the apparent endemic descent of common sense in these modern times i
thought it would be nice for the denizens of MFW to read a book
review. it isn't the same as an interview with the author on a
kick-ass fake news show, but it'll do for now.
Reviewed: The Evolution-Creation Struggle by Michael Ruse
30 July 2005, New Scientist
Karen Armstrong reviewing Michael Ruse's "The Evolution-Creation
Struggle", Harvard University Press, 2005
THE clash between those who adhere to the scientific theory of
evolution and those who believe that the biblical story of the six-day
creation is literally true is a struggle between two religions. So
concludes Michael Ruse in his accessible, skilfully written book.
Since the Enlightenment, he says, scientists have offered up an
alternative vision of the nature of reality, and those among them who
are most opposed to religion can proselytise with as much zeal as an
evangelical Christian.
For Richard Dawkins, contemplation of the natural world through the
eyes of science is a religious experience, providing the same
"spine-shivering, breath-catching awe - almost worship" as cultivated
by the great religions.
Yet Dawkins regards faith as one of the world's great evils,
"comparable to the smallpox virus but harder to eradicate". He defines
faith as "belief that isn't based on evidence", whereas science is
based upon verifiable facts. Even those who would not go as far in
their condemnation of religion would probably agree that modern
science has undermined the conventional notion of God. There is a
widespread, popular conviction that science and religion are
diametrically opposed, and that science has rendered most religious
truth frankly incredible.
But this conviction is based on an erroneous assumption: that faith is
synonymous with belief, and that to be religious, people must accept
certain credal propositions. This is a relatively recent development,
one that has arisen since the Enlightenment, and then only in the
west. Originally the Middle English word beleven meant "to love". The
Latin credo probably derives from co do: "I give my heart". Faith was
therefore not belief but commitment. Even Martin Luther, who taught
that human beings were justified by faith, did not define faith by
belief; he had in fact very little time for dogma and creeds. Faith
was a heroic cultivation of trust in the idea that, against all the
evidence to the contrary, life had some ultimate, though ineffable,
meaning and value.
Religion is not about thinking, but doing things that change you at a
profound level熱ost of the great religions have had no interest in
metaphysical doctrines. Religion is not about thinking things but
about doing things that change you at a profound level. At best,
theology is regarded as a kind of poetry about matters that must, by
their very nature, elude definition. The Koran calls theological
speculation zanna, self-indulgent guesswork about matters that nobody
can prove one way or the other. Every single verse, every story told
in the Koran is called an aya, a parable, because it is only possible
to talk about the indescribable God in terms of signs and symbols.
Hence, Darwin and his evolutionary theory has raised scarcely a ripple
of concern in the Muslim world.
But the biblical creation myths were never intended as definitive
dogma either. The first chapter of Genesis, which was almost certainly
written by a priestly author (often referred to as "P") who had been
deported to Babylonia after the destruction of Jerusalem by King
Nebuchadnezzar in 586 BC, became the most famous creation account, but
it was not the only one. The biblical editors placed it right next to
another story that contradicts P's in several important respects.
Unlike P's creation story, most Middle Eastern cosmologies were
extremely violent. The Babylonian god Marduk, for example, slew the
divine sea monster Tiamat after a bloody battle and split her carcass
in two, like a giant shellfish, to create heaven and earth. Such
stories expressed the tragic - not to say Darwinian - insight that
life, creativity and survival depended upon the destruction of others.
Genesis chapter 1 was revolutionary in omitting all the violence. It
imagines Yahweh - the God of the Bible - summoning all things into
being with a mere word of command. P was telling the exiled Jews that
their God was far more powerful than Marduk. His calm creation was a
healing vision of order, designed to console the traumatised
deportees.
Cosmologies were originally therapeutic in function. In the ancient
world, a creation story was often chanted at the start of a new
enterprise or at a moment of crisis. These stories were thus more than
history. Nobody, not even the gods, knew what had happened at the
beginning of time. "Who then knows whence this creation has arisen?"
asks the inspired poet of the Hindu Rig Veda. "Only he who is its
overseer in highest heaven knows - or perhaps he does not know!"
Until the advent of the modern period, nobody would have regarded the
six-day creation story as a literal, historical account. In the 16th
century, for example, after the Jews had been expelled from Spain, the
Kabbalist Isaac Luria evolved an entirely new creation myth that had
nothing in common with P's story, but was full of explosions and false
starts. Far from being reviled for contradicting scripture, Lurianic
Kabbalah became a mass movement, for it expressed the pain and
bewilderment of being Jewish at that time and showed scattered,
persecuted people how they could rebuild their world.
In the pre-modern world, it was generally understood that there were
two ways of arriving at truth. Plato called them mythos and logos.
Neither was superior to the other. Logos (reason; science) was exact,
practical and essential to human life. To be effective, it had to
correspond to external reality. Myth expressed the more elusive,
puzzling aspects of human experience. It has often been called a
primitive form of psychology, which helped people negotiate their
inner world. A mythical story, such as a cosmology, described
something that had happened once, but also happened all the time. A
myth was essentially a programme of action. Unless you put it into
practice, you could not judge its truth.
Myth could not help you create efficient technology or run your
society. But logos had its limits too. If you became a refugee or
witnessed a terrible natural catastrophe, you did not simply want a
logical explanation; you also wanted myth to show you how to manage
your grief. With the advent of our scientific modernity, however,
logos achieved such spectacular results that myth was discredited, and
now, in popular parlance a myth is something that did not happen, that
is untrue. But some religious people also began to read religious
myths as though they were logos.
The conflict between science and faith has thus been based on a
misunderstanding of the nature of scriptural discourse. Many people,
including those who are religious, find it difficult to think
mythically, because our education and society is fuelled entirely by
logos. This has made religion impossible for many people in the west,
and it could be argued that much of the stridency of Christian
fundamentalism is based on a buried fear of creeping unbelief.
In the pre-modern world, it was considered dangerous to mix mythos and
logos, because each had a different sphere of competence. Much of the
heat could be taken out of the evolution versus creation struggle if
it were admitted that to read the first chapter of Genesis as though
it were an exact account of the origins of life is not only bad
science; it is also bad religion.
From issue 2510 of New Scientist magazine, 30 July 2005, page 42
Axel:
anyone out there hear the NPS program by that one Rabbi dude? that was
interesting. the old testament's punishment of killing a fetus doesn't
seem to carry anything like the weight of killing a human being. i'd
agree. also the opinion on the judaic take on homosexuality.. though
that may be left to interpretation. i don't particularly care for it
(except for some genuinely hot lesbianism, of course) but i do find
the irresponsible and narcissistic unrestrained plunge into hedonism a
bit offensive. let's face it, what religion would support a subset of
its population that doesn't reproduce?
why is it so difficult to allow for the universe that God created to
have been created over a long period of time? why can't every species
that emerges from this universe have its own communion with the
divine, actively seeking not only its own survival but also its own
evolutionary development? can't we all be divine creatures? why do
people insist on a puerile version of an anthropomorphic god of
punishment and reward? i think that's a failure of individual
evolution. i see our species, right here in america, taking a
disturbing turn towards devolution.
--
"... the street term for it would be 'Byxmyndig' *legal to get in
pants* (it's true)" - Dukeman, in AGHL
David Cohen
DZ wrote:
> Question is, why all those people who never wrote a single paper on
> evolution theory keep writing books on it?
God told them to.
Duh.
David
ATP*
"Axel of the North!" <ax...@ypo.net> wrote in message
news:4331a62...@netnews.worldnet.att.net...
> Hey kids! Whoa whoa whoa whoa whoa! Ughhh...
>
> After seeing the Daily Show's weeklong "Evolution Schmevolution" and
> the apparent endemic descent of common sense in these modern times i
> thought it would be nice for the denizens of MFW to read a book
> review. it isn't the same as an interview with the author on a
> kick-ass fake news show, but it'll do for now.
>
> Reviewed: The Evolution-Creation Struggle by Michael Ruse
> 30 July 2005, New Scientist
>
> Karen Armstrong reviewing Michael Ruse's "The Evolution-Creation
> Struggle", Harvard University Press, 2005
> The Koran calls theological
> speculation zanna, self-indulgent guesswork about matters that nobody
> can prove one way or the other. Every single verse, every story told
> in the Koran is called an aya, a parable, because it is only possible
> to talk about the indescribable God in terms of signs and symbols.
> Hence, Darwin and his evolutionary theory has raised scarcely a ripple
> of concern in the Muslim world.
>
That must explain their tolerance for diverse interpretations of God's word
and general open-mindedness......................
Dr. Dickie
"Joe Humble" <joeh...@earthlink.net> wrote in message
news:4hh6j1pl2dakfdkeo...@4ax.com...
> On Wed, 21 Sep 2005 18:47:41 GMT, ax...@ypo.net (Axel of the North!)
> wrote:
>
> >Hey kids! Whoa whoa whoa whoa whoa! Ughhh...
> >
> >After seeing the Daily Show's weeklong "Evolution Schmevolution" and
> >the apparent endemic descent of common sense in these modern times i
> >thought it would be nice for the denizens of MFW to read a book
> >review. it isn't the same as an interview with the author on a
> >kick-ass fake news show, but it'll do for now.
> >
> >Reviewed: The Evolution-Creation Struggle by Michael Ruse
> >30 July 2005, New Scientist
> >
>
> two distinctly different things and suggesting there is a struggle
> between them is like suggesting there is a struggle between paper and
> sheetrock.
>
Actually there is; albeit, it is one-sided. Religion railies against science
all the time (the fanatics at least). Reason, science has removed most of
the fear based tactics that religion used to keep the masses in line (that
is, supporting the church). Science did this by providing sound, scientific
reasons for natural phenomena that religion once used to demonstrate, "The
wrath of God!!"
> Holy crap, that article goes on WAY too long.
>
>
If you cannot dazzle them with fact, bore them into submission ;-)
--
Dr. Dickie
Skepticult member in good standing #394-00596-438
Poking kooks with a pointy stick
Dr. Dickie
"Dr. Dickie" <Dr_D...@chembench.com> wrote in message
news:1127474104.cdb4ce13c76ab5fc9dd72bf1b51f93dc@teranews...
>
> "Joe Humble" <joeh...@earthlink.net> wrote in message
> news:4hh6j1pl2dakfdkeo...@4ax.com...
> > On Wed, 21 Sep 2005 18:47:41 GMT, ax...@ypo.net (Axel of the North!)
> > wrote:
> >
> > >Hey kids! Whoa whoa whoa whoa whoa! Ughhh...
> > >
> > >After seeing the Daily Show's weeklong "Evolution Schmevolution" and
> > >the apparent endemic descent of common sense in these modern times i
> > >thought it would be nice for the denizens of MFW to read a book
> > >review. it isn't the same as an interview with the author on a
> > >kick-ass fake news show, but it'll do for now.
> > >
> > >Reviewed: The Evolution-Creation Struggle by Michael Ruse
> > >30 July 2005, New Scientist
> > >
> >
> > There is no inherent struggle between science and the Bible. They are
> > two distinctly different things and suggesting there is a struggle
> > between them is like suggesting there is a struggle between paper and
> > sheetrock.
> >
>
> Actually there is; albeit, it is one-sided. Religion railies against
science
> all the time (the fanatics at least). Reason, science has removed most of
> the fear based tactics that religion used to keep the masses in line (that
> is, supporting the church). Science did this by providing sound,
scientific
> reasons for natural phenomena that religion once used to demonstrate, "The
> wrath of God!!"
>
There is even a great example, the hurricanes that are currently plaguing
the USA.
Many years ago (the dark ages, when religion was king), we would have
thought that these were the result of gods wrath.
Today, thanks to science, we know they are caused by G.W. Bush (just ask the
news media!).
--
------
Dr. Dickie
"Let be be finale of seem.
The only emperor is the emperor of ice-cream."
-- Wallace Stevens
The Bill Rodgers
<Dr_D...@chembench.com> wrote:
>Today, thanks to science, we know they are caused by G.W. Bush (just ask the
>news media!).
God is angry with the chimp for using his name to justify killing
thousands of people.
TBR
"As democracy is perfected, the office of president represents, more and
more closely, the inner soul of the people. On some great and glorious day
the plain folks of the land will reach their heart's desire at last and the
White House will be adorned by a downright moron."
H.L. Mencken (1880 - 1956)
"Anyone with degrees from Yale and Harvard is presumed to be intelligent,
but George W. Bush has managed to overcome that presumption."
ATP*
"DZ" <3108*3@190*7630*056.1997*4688*0.30*38*0.2*59*1.9*29*8> wrote in
message news:21...@1281131942.1648826963.24725.8767.21074...
> Dr. Dickie <Dr_D...@chembench.com> wrote:
>> "Joe Humble" <joeh...@earthlink.net> wrote
>>> ax...@ypo.net (Axel of the North!) wrote:
>>>
>>> >Hey kids! Whoa whoa whoa whoa whoa! Ughhh...
>>>
>>
>> If you cannot dazzle them with fact, bore them into submission ;-)
>
> USSR-wide student competition for the best paper and presentation on
> philosophy of science, where my good friend and me as co-authors took
> the second place. I was presenting. A chemistry guy from the Republic
> of Georgia took the first place. At first I had some sense of
> accomplishment, but shortly after the event become totally
> disillusioned. I now reduce the whole philosophy of science to a bunch
> of stupid rules, like: there isn't good without bad; no smoke without
> fire; every coin has the opposite side, etc. Mix in some fluff and
> you're ready to write books on Intelligent Design.
>
> IMO, "science philosophers", who never bothered to do any actual
> science, are a bunch of useless smoke blowers who only get in
> the way :-)
That was the way it was in the early stages of modern Western scientific
thought. Math, philosophy and science were not developed as discrete
subjects. There was obviously more room at the time for some pretty fanciful
constructs that could not be proven or disproven. Nevertheless, some of
those philosopher's names are still prominently represented in math texts.
ATP*
"DZ" <37...@3251832624.2417511610.25683.292.21243> wrote in message
news:12...@132422371.304495619.29066.8527.4176...
> ATP* <eug...@ogedin.com> wrote:
> philosopher. BTW, modern theory of evolution is largely applied
> mathematics. But philosophers of science who aren't primarily
> scientists don't have anything useful to tell us. They only spread the
> kind of drivel that makes it hard to teach evolution in schools and as
> a consequence hinder the progress of genetic and medical research.
Agreed. At this point, there is not much need for that type of mental
masturbation.
Dr_Dickie
--
Dr. Dickie
Skepticult member in good standing #394-00596-438
Poking kooks with a pointy stick.
"The most exciting phrase to hear in science, the one that heralds new
discoveries,
is not 'Eureka!' ('I found it!'), but rather 'hmm....that's funny...'"
- Isaac Asimov
"DZ" <3108*3@190*7630*056.1997*4688*0.30*38*0.2*59*1.9*29*8> wrote in
message news:21...@1281131942.1648826963.24725.8767.21074...
> Dr. Dickie <Dr_D...@chembench.com> wrote:
> > "Joe Humble" <joeh...@earthlink.net> wrote
> >> ax...@ypo.net (Axel of the North!) wrote:
> >>
> >> >Hey kids! Whoa whoa whoa whoa whoa! Ughhh...
> >>
> >> Holy crap, that article goes on WAY too long.
> >
> > If you cannot dazzle them with fact, bore them into submission ;-)
>
> Long time ago Novosibirsk State University had been holding a
> USSR-wide student competition for the best paper and presentation on
> philosophy of science, where my good friend and me as co-authors took
> the second place. I was presenting. A chemistry guy from the Republic
> of Georgia took the first place. At first I had some sense of
> accomplishment, but shortly after the event become totally
> disillusioned. I now reduce the whole philosophy of science to a bunch
> of stupid rules, like: there isn't good without bad; no smoke without
> fire; every coin has the opposite side, etc. Mix in some fluff and
> you're ready to write books on Intelligent Design.
>
> IMO, "science philosophers", who never bothered to do any actual
> science, are a bunch of useless smoke blowers who only get in
> the way :-)
Well, I don't know. Have you read any of Poppin's stuff?
That is the only philosophy of science I am familiar with.
BTW, intelligent design is not a "theory." In science, the term theory
refers to a set of rules with a "large body of evidence" supporting it.
Evolution is a theory, intelligent design is an interesting concept (and I
actually want to believe it is true, but that is a belief, NOT science--I am
allowed my personal beliefs).
--
Dr. Dickie
Skepticult member in good standing #394-00596-438
Poking kooks with a pointy stick.
"The most exciting phrase to hear in science, the one that heralds new
discoveries,
is not 'Eureka!' ('I found it!'), but rather 'hmm....that's funny...'"
- Isaac Asimov
Dr_Dickie
"Dr_Dickie" <Dr_D...@chembench.com> wrote in message
news:1127735908.d1b1ad43465aed9e6d75ebebe105b9a0@teranews...
Sorry, I mean Popper, Karl Popper.
Dr_Dickie
"DZ" <24...@102492322.187963792.32497.29121.5191> wrote in message
news:2...@223924585.15058232.25711.23425.12768...
> Dr_Dickie <Dr_D...@chembench.com> wrote:
> > "Dr_Dickie" <Dr_D...@chembench.com> wrote:
> scientific discovery", in the 80s. How could I not? - after all, I'm
> an accomplished philosopher of science ;-) BTW, there Popper opined
> that there is no such thing as positive evidence, so I'm not sure how
> you can talk about Popper and refer to some nebulous "large body of
> evidence supporting" something, in the same paragraph. Too much
> Popper, too badly read, Professor :-)
Hey, I said I like Popper, I did not say all of the rest of science goes by
Popper's reasoning (I do agree with him though; although, I think he would
agree that when scientists say that something is supported by evidence, they
mean that it has been tested and a non-negative result has been found.)
>
> Light-hearted discussion of science as opposed to pompous preaching
> along the lines of "how y'all should do science" could be
> interesting. I've read with interest some science philosophy papers by
> a Bayesian statistician IJ Good. But those are to an extent like
> probability puzzles and he's really "Good" at that, e.g.
>
> "The hypothesis that all crows are black is the same as that all
> non-black things are not crows, and this is supported by the
> observation of a white shoe". True or False?
>
> (answer - the white shoe is a red herring).
Could be fun for mental conditioning (since I work in Alzheimer's research,
this is never far from my thoughts); however, I am much more interested in
the philosophy that goes into the formation of the scientific method. I
should dust off my Popper and give it another read this winter.
Any suggestions for someone else? How about Popper's protégé, David Miller,
any good?
Hugh Beyer
news:1127820053.fe6e7e6c7079a6c04c9adc93b4e95ad0@teranews:
>> >> Well, I don't know. Have you read any of Poppin's stuff?
>> >> That is the only philosophy of science I am familiar with.
>> >> BTW, intelligent design is not a "theory." In science, the term
>> >> theory refers to a set of rules with a "large body of evidence"
>> >> supporting it. Evolution is a theory, intelligent design is an
>> >> interesting concept (and I actually want to believe it is true, but
>> >> that is a belief, NOT science--I am allowed my personal beliefs).
>> >>
>> > Sorry, I mean Popper, Karl Popper.
>>
>> I don't think ID is a theory either. I've read "The logic of
>> scientific discovery", in the 80s. How could I not? - after all, I'm
>> an accomplished philosopher of science ;-) BTW, there Popper opined
>> that there is no such thing as positive evidence, so I'm not sure how
>> you can talk about Popper and refer to some nebulous "large body of
>> evidence supporting" something, in the same paragraph. Too much
>> Popper, too badly read, Professor :-)
>
> Hey, I said I like Popper, I did not say all of the rest of science goes
> by Popper's reasoning (I do agree with him though; although, I think he
> would agree that when scientists say that something is supported by
> evidence, they mean that it has been tested and a non-negative result
> has been found.)
>
I think the point here is that a theory can never be proved, only disproved.
So Einstein's theory of relativity hasn't been proved, despite the large
body of evidence supporting it; someone could show up tomorrow with an
experimental result contradicting that predicted by relativity. Relativity
would then need to be replaced with a better model.
Which has interesting implications for Intelligent Design. Even if we could
figure out how to turn ID into a scientifically testable hypothesis, we
still couldn't prove the existance of the Designer--we could only show that
we had failed to disprove it.
Hugh
--
Exercise is a dirty word. Whenever I hear it, I wash my mouth out with
chocolate. ("Ladi")
Dr. Dickie
"Hugh Beyer" <beyer...@acm.org> wrote in message
news:Xns96DE81E7C30C...@130.81.64.196...
That is the thesis of Popper's work (and in fact, that is the scientific
method--relativity did just that to Newtonian physics); however, Popper does
conclude that the only theories worth considering are those that have a
large amount of information and have produced non-negative results when
tested (the more the testing the better).
Also, it does not necessarilly need to be replace, just better defined as to
what the constraints of the theory are (as NO scientific theory works 100%
of the time).
> Which has interesting implications for Intelligent Design. Even if we
could
> figure out how to turn ID into a scientifically testable hypothesis, we
> still couldn't prove the existance of the Designer--we could only show
that
> we had failed to disprove it.
>
> Hugh
>
Right, which is why it is not scientific. The heart of Popper's work was
that something which cannot POSSIBLY give a negative result is not part of
science.
Faith, on the other hand, incorporates all else. Kinda makes for a nice neat
package. Problems, of course, arise when the two are confused.
Personally, bright shiny objects tend to do that to me ;-)
WillBrink
Hugh Beyer <beyer...@acm.org> wrote:
Not actually true. The major points of Einstein零 theory on relativity
have been supported by physical and mathematical evidence.
> despite the large
> body of evidence supporting it; someone could show up tomorrow with an
> experimental result contradicting that predicted by relativity. Relativity
> would then need to be replaced with a better model.
See above. So far, above the quantum level, relativity appears to hold
true. At the quantum level, it does not and there may be an entirely
different set of rules (string theory, and so on) and what is now the
holy grail of physics is a unified theory that ties the two together.
>
> Which has interesting implications for Intelligent Design.
Errr, no it does not actually.
> Even if we could
> figure out how to turn ID into a scientifically testable hypothesis, we
> still couldn't prove the existance of the Designer
I don't see why. If there was a testable hypothesis for ID, then you
could show there was at least *some* designer, ergo, that something with
intelligence had an active hand in shaping the universe.
>-we could only show that
> we had failed to disprove it.
It's a start, and some times the negative result you get in science is
just as telling as the pos result.
>
> Hugh
--
Will Brink @ http://www.brinkzone.com/
Dr. Dickie
"WillBrink" <WillBrink*NOSPAM*@Comcast.net> wrote in message
news:WillBrink*NOSPAM*-67DBC6.135...@comcast.dca.giganews.com...
No, so far there have simply been non-negative results. I, as I believe
Popper did, take that be supportive, but the point is it cannot be proven,
only not yet disproven.
>
> > despite the large
> > body of evidence supporting it; someone could show up tomorrow with an
> > experimental result contradicting that predicted by relativity.
Relativity
> > would then need to be replaced with a better model.
>
> See above. So far, above the quantum level, relativity appears to hold
> true. At the quantum level, it does not and there may be an entirely
> different set of rules (string theory, and so on) and what is now the
> holy grail of physics is a unified theory that ties the two together.
>
> >
> > Which has interesting implications for Intelligent Design.
>
> Errr, no it does not actually.
>
> > Even if we could
> > figure out how to turn ID into a scientifically testable hypothesis, we
> > still couldn't prove the existance of the Designer
>
> I don't see why. If there was a testable hypothesis for ID, then you
> could show there was at least *some* designer, ergo, that something with
> intelligence had an active hand in shaping the universe.
>
But that is the rub, how do you design an experiment that COULD produce
negative results without testing the designer.
> >-we could only show that
> > we had failed to disprove it.
>
> It's a start, and some times the negative result you get in science is
> just as telling as the pos result.
>
Oh, not sometimes, everytime. A negative result is essential.
If you do an experiment and get a positive result, you know nothing. For all
you know, no matter what you do you will get a positive result.
Do it again, and get a negative result, NOW you should now have learned
something about how it works.
WillBrink
"Dr. Dickie" <Dr_D...@chembench.com> wrote:
That is not true. For example, Einstein零 theory of the relative
differences in time between two objects has been physically proven via
two atomic clocks, and so on.
> I, as I believe
> Popper did, take that be supportive, but the point is it cannot be proven,
> only not yet disproven.
Einstein零 theories cannot be proven only disproven? Is that what you
are saying?
>
> >
> > > despite the large
> > > body of evidence supporting it; someone could show up tomorrow with an
> > > experimental result contradicting that predicted by relativity.
> Relativity
> > > would then need to be replaced with a better model.
> >
> > See above. So far, above the quantum level, relativity appears to hold
> > true. At the quantum level, it does not and there may be an entirely
> > different set of rules (string theory, and so on) and what is now the
> > holy grail of physics is a unified theory that ties the two together.
> >
> > >
> > > Which has interesting implications for Intelligent Design.
> >
> > Errr, no it does not actually.
> >
> > > Even if we could
> > > figure out how to turn ID into a scientifically testable hypothesis, we
> > > still couldn't prove the existance of the Designer
> >
> > I don't see why. If there was a testable hypothesis for ID, then you
> > could show there was at least *some* designer, ergo, that something with
> > intelligence had an active hand in shaping the universe.
> >
> But that is the rub, how do you design an experiment that COULD produce
> negative results without testing the designer.
You can't due to the fact there is no designer, but that's not my point.
Hugh Beyer
news:WillBrink*NOSPAM*-1E8449.142...@comcast.dca.giganews.com:
> In article <1127844529.76475a6689bde2fa0ea19c49feb52768@teranews>,
> "Dr. Dickie" <Dr_D...@chembench.com> wrote:
>
>> "WillBrink" <WillBrink*NOSPAM*@Comcast.net> wrote in message
>> news:WillBrink*NOSPAM*-67DBC6.13514227092005
You're missing the point here. All that experiment proved was that if you
take two atomic locks and accelerate one around the earth a few dozen
times, when you bring them back together they are out of synch to the same
degree that Relativity predicts they will be. That doesn't prove
Relativity, it only fails to disprove the theory in this case.
There's actually a cottage industry in physics in coming up with
alternative theories to relativity that explain the same experimental
results. One day, one of them might turn out to be a better fit to the
data than relativity.
Point is the best we can say is that relativity has shown to predict
reality in enough different situations that we can be quite confident the
model is a close match to reality, and that even if it turns out to be a
poor match under some yet-to-be-discovered conditions, it will still be
useful in a wide range of situations--just as Newton's laws, though wrong,
are still useful.
>
>> I, as I believe
>> Popper did, take that be supportive, but the point is it cannot be
>> proven, only not yet disproven.
>
> Einstein零 theories cannot be proven only disproven? Is that what you
> are saying?
Yep. Can't prove a theory. See above.
Which is why saying "evolution is only a theory" is nonsense. They're
*all* only theories. Has nothing to do with truth or utility.
>>
>> >
>> > > despite the large
>> > > body of evidence supporting it; someone could show up tomorrow with
>> > > an experimental result contradicting that predicted by relativity.
>> > > Relativity would then need to be replaced with a better model.
>> >
>> > See above. So far, above the quantum level, relativity appears to
>> > hold true. At the quantum level, it does not and there may be an
>> > entirely different set of rules (string theory, and so on) and what
>> > is now the holy grail of physics is a unified theory that ties the
>> > two together.
>> >
>> > >
>> > > Which has interesting implications for Intelligent Design.
>> >
>> > Errr, no it does not actually.
>> >
>> > > Even if we could
>> > > figure out how to turn ID into a scientifically testable
>> > > hypothesis, we still couldn't prove the existance of the Designer
>> >
>> > I don't see why. If there was a testable hypothesis for ID, then you
>> > could show there was at least *some* designer, ergo, that something
>> > with intelligence had an active hand in shaping the universe.
Because all you can test is a theory's effects--you can't test the theory
itself. Suppose, for example, I claim that any intelligence capable of
producing a universe that looks like a mechanism isn't going to be able to
resist signing His work. That's a testable hypothesis--somewhere out there
is a Cosmic Signature that we can find. Now suppose I find that viewed
from a certain vantage point, there's a cluster of stars that spell out
"Kilroy was here" across the face of a galaxy. Have I proven ID? I have
not. An alternative theory explaining this data is that if you take all
possible vantage points and all possible galaxies, it's not too surprising
to find one which seems to spell out a message.
It's not a wonderful example because the test is too hard to execute--but
if you could show that nowhere in the universe is any sort of signature,
you would have proven that there is no Intelligent Designer of the sort
that needs to sign His work.
Charles
wrote:
But what if they are both sent around the world a few dozen times, but
in opposite directions. Would that double the time difference or make
the clocks both record identical times?
Would it not be simpler to use Occam's principle of parsimony (Occam's
Razor), and agree the simplest and most likely conclusion, accepting
that we are all terminally fucked and will never know what this
bizarre existence is all about?
Dr. Dickie
> WillBrink <WillBrink*NOSPAM*@Comcast.net> wrote in
> news:WillBrink*NOSPAM*-1E8449.142...@comcast.dca.giganews.com:
>
> > In article <1127844529.76475a6689bde2fa0ea19c49feb52768@teranews>,
> > "Dr. Dickie" <Dr_D...@chembench.com> wrote:
> >
> >> No, so far there have simply been non-negative results.
> >
> > That is not true. For example, Einstein零 theory of the relative
> > differences in time between two objects has been physically proven via
> > two atomic clocks, and so on.
>
> You're missing the point here. All that experiment proved was that if you
> take two atomic locks and accelerate one around the earth a few dozen
> times, when you bring them back together they are out of synch to the same
> degree that Relativity predicts they will be. That doesn't prove
> Relativity, it only fails to disprove the theory in this case.
>
Right, remember Newtonian physics was absoulutly true (and still is to about
3 decimal places as long as you are sub 5% of the speed of light) and was
shown to give the correct answer time and time again. Then Einstien came
along and showed that it was not correct.
What you have to keep in mind here is, we are modeling things we really do
not completely understand. In chemistry, there are serveral bonding
theories. Are any of them correct! No. However, each on can yeild (that is,
predict correctly) a correct answer when used within the constraints of the
theory.
Correct. There is no way to get a negative result when searching for
Intelligent design (although I do think Hank comes close). You can always
say, "Well, we just don't understand the intelligence behind the design."
Sort of the, " No one can know the mind of God," answer I used to love to
get from the nuns in school. When I head that, I knew that I had won the
argument ;-)
Dr. Dickie
"Charles" <ch...@charles.net> wrote in message
news:97gjj1pk0umatmhhc...@4ax.com...
Boy, the "extraterrestrial alien nuts" HATE Occam's razor.
Isn't is simpler that drunk British blokes out for some fun stamped that
design into the wheat?
Oh NO. Highly intelligent space aliens traveled (thus disproving relativity)
to earth, but at the last minute decided that we could not handle their
message (why do I hear Jack Nickason in my head) straight out. So, they
tramped their message into the wheat knowing that when we were ready to
receive it we would understand it (kinda of 2001 A Space Odyssey meets Field
of Dreams).
Dr. Dickie
"DZ" <95...@1019220230.233293419.25904.11480.26912> wrote in message
news:5431@1202223624.3016026462.15725.2935.21894...
> Dr_Dickie <Dr_D...@chembench.com> wrote:
> > Hey, I said I like Popper, I did not say all of the rest of science
> > goes by Popper's reasoning (I do agree with him though; although, I
> > think he would agree that when scientists say that something is
> > supported by evidence, they mean that it has been tested and a
> > non-negative result has been found.)
>
>
> So what happens when a "positive" result is found? Does that increase
> the evidence toward a hypothesis, or does it decrease it?
>
> Let me assure you that a result in agreement with a hypothesis (i.e. a
> confirmatory, or positive result) in fact DECREASES the evidence for
> that hypothesis! How about that? :-)
>
> Suppose our hypothesis is that all gun owners in a given country are
> criminals. If we observe one or more gun-owners who are also
> criminals, does that support our hypothesis? No, it REFUTES it!
>
> A "good" philosopher of science like, IJ Good, can easily defend such
> point of view.
>
> We are brought blindfolded into either of two imaginary
> countries. Let's call them "US" and "UK", for simplicity.
>
> We're presented with the information that -
>
> 1) In the "UK" there are 2 criminal gun owners, notorious Hadov and
> his butler (posession of guns there is outlawed), no any other gun
> owners, and about 100mln of other people.
>
> 2) In the "US", there are 50mln criminal gun owners, one non-criminal
> gun owner (Dally), and lets say 100mln of other people who have no
> guns whatsoever (like Cohen).
>
> We're presented with a randomly chosen person who appears to be a
> criminal gun owner. First of all, this means - chances are, we're in
> the US. But in that country not every gun owner is a criminal.
> Therefore this POSITIVE confirmatory observation
> (gun owner = criminal) in fact weakens the evidence for the hypothesis
> that all gun owners in the country are criminal.
>
> QED
No, a gun owner that is a criminal does not weaken the hypothesis that not
EVERY gun owner is a criminal, in fact it supports it (it is a non-negative
result--since not every gun owner is a criminal leaves the door open for
some gun owners are criminal).
Remember, Popper says that the only hypothesis's that are worth considering
are the ones with the a huge amount of information and testable. The
hypothesis that not every gun owner is a criminal, is very low in
information.
In addition, you are dealing with statistics,and the statistics of small
numbers (n=1), as I am sure you very well know, is not even worth discussing
;-0
And remember, it is "A body of evidence" that makes a theory accepted, not a
single observation.
We are modeling the unknown, and attempting to explain the un-explainable. I
think it was Fermi that said, "If you think you understand quantum
mechanics, you don't."
Those of us that know how ignorant we are, are fighting for every morsel of
understanding (sometimes we are too aggressive in accepting something before
there really is enough know--we are human).
Charles
<Dr_D...@chembench.com> wrote:
>
>"Charles" <ch...@charles.net> wrote in message
>news:97gjj1pk0umatmhhc...@4ax.com...
>> On Tue, 27 Sep 2005 21:36:04 GMT, Hugh Beyer <beyer...@acm.org>
>> wrote:
>>
>> >It's not a wonderful example because the test is too hard to execute--but
>> >if you could show that nowhere in the universe is any sort of signature,
>> >you would have proven that there is no Intelligent Designer of the sort
>> >that needs to sign His work.
>>
>> Would it not be simpler to use Occam's principle of parsimony (Occam's
>> Razor), and agree the simplest and most likely conclusion, accepting
>> that we are all terminally fucked and will never know what this
>> bizarre existence is all about?
>
>Boy, the "extraterrestrial alien nuts" HATE Occam's razor.
>Isn't is simpler that drunk British blokes out for some fun stamped that
>design into the wheat?
That works for me.
>Oh NO. Highly intelligent space aliens traveled (thus disproving relativity)
>to earth, but at the last minute decided that we could not handle their
>message (why do I hear Jack Nickason in my head) straight out. So, they
>tramped their message into the wheat knowing that when we were ready to
>receive it we would understand it (kinda of 2001 A Space Odyssey meets Field
>of Dreams).
Sounds somewhat spaced out! ;o)
Hugh Beyer
520626057.973829628.6348.19868.22778:
> Dr. Dickie <Dr_D...@chembench.com> wrote:
> It does weaken it, if you consider the odds of the hypothesis and its
> complement through the formal computation.
>
>> Remember, Popper says that the only hypothesis's that are worth
>> considering are the ones with the a huge amount of information and
>> testable. The hypothesis that not every gun owner is a criminal, is
>> very low in information.
>
> It really doesn't matter what Popper thinks is WORTH considering. If
> philosophy of science only "works" for a subset of hypotheses that are
> worth considering, then screw philosophy of science! :-)
>
> (I guess that is what I was saying right from the beginning)
>
>> In addition, you are dealing with statistics,and the statistics of
>> small numbers (n=1), as I am sure you very well know, is not even
>> worth discussing ;-0
>
> This is wrong. In statistics, every new observation is shifting the
> existing odds even if by a small amount. This is the basis for
> sequential inference. A decision is made when a single new observation
> pushes the odds above a certain threshold. In the example considered,
> any new observation of a criminal gun owner is shifting the odds
> towards the alternative "not every gun owner is a criminal". If you
> actually compute the odds, the evidence against the hypothesis, as I
> stated it, is overwhelming.
>
> It is more illuminating if the countries are replaced by two mutually
> exclusive universes (you can't be in both at the same time), with
> distinctive fixed rules.
>
>> And remember, it is "A body of evidence" that makes a theory
>> accepted, not a single observation.
>
> You are welcome to modify that thought experiment to be allowed to
> keep sampling people. When another gun owner criminal is sampled, the
> evidence against the hypothesis (gun owner = criminal) will become
> even stonger.
>
The way you set up the situation isn't entirely clear. I think you meant:
You select a person at random for the country you're in, and you don't
know what country you're in. That random person is a criminal gun owner.
Therefore you've raised the probablility that you're in the US (because
you saw a gun owner at all) and therefore the probablility that in this
country, not all gun owners are criminal.
It's cute, but it's not demonstrating what you claim. You're conflating
two different questions--the probability that you are in the US with the
probability that the statement "all gun owners are criminal" is true. Your
sample of one is raising the probability of the first and you've rigged it
so that the first implies the second FOR THIS COUNTRY.
DZ
Chances of being in one of the countries given the data is an integral
part of the calculation of the odds in favor of the hypothesis, the
way I stated it, rather than a conflation of two issues.
> Your sample of one is raising the probability of the first and
> you've rigged it so that the first implies the second FOR THIS
> COUNTRY.
That's exactly how I formulated the hypothesis:
"...our hypothesis is that all gun owners IN A GIVEN COUNTRY are
criminals".
Imagine two possible universes "UK" vs. "US" and a mathematical theory
that predicts two distinct sets of natural laws for these two
worlds. However, only one of the two worlds can be realized, with
equal probability. You are to believe the theory. You are in one of
the two possible worlds but don't know in which one.
In this setup, it doesn't make sense to calculate the probability
averaged over two mutually exclusive realizations.
DZ
Dr_Dickie
--
Dr. Dickie
Skepticult member in good standing #394-00596-438
Poking kooks with a pointy stick.
"The most exciting phrase to hear in science, the one that heralds new
discoveries,
is not 'Eureka!' ('I found it!'), but rather 'hmm....that's funny...'"
- Isaac Asimov
"DZ" <14...@2853024297.86234330.703.20658.31447> wrote in message
news:20...@520626057.973829628.6348.19868.22778...
> Dr. Dickie <Dr_D...@chembench.com> wrote:
> > "DZ" wrote:
> It does weaken it, if you consider the odds of the hypothesis and its
> complement through the formal computation.
Weaken is very open term. It does not violate the premise, so how can it
weaken it?
>
> > Remember, Popper says that the only hypothesis's that are worth
> > considering are the ones with the a huge amount of information and
> > testable. The hypothesis that not every gun owner is a criminal, is
> > very low in information.
>
> It really doesn't matter what Popper thinks is WORTH considering. If
> philosophy of science only "works" for a subset of hypotheses that are
> worth considering, then screw philosophy of science! :-)
>
We are not talking about what Popper considers worth considering, we are
talking about what science should be considering.
> (I guess that is what I was saying right from the beginning)
>
> > In addition, you are dealing with statistics,and the statistics of
> > small numbers (n=1), as I am sure you very well know, is not even
> > worth discussing ;-0
>
> This is wrong. In statistics, every new observation is shifting the
> existing odds even if by a small amount.
No, no matter how many times you filp a coin (a perfect coin, perfect flip),
the odds are still 50:50 you will get tails.
>This is the basis for
> sequential inference. A decision is made when a single new observation
> pushes the odds above a certain threshold. In the example considered,
> any new observation of a criminal gun owner is shifting the odds
> towards the alternative "not every gun owner is a criminal". If you
> actually compute the odds, the evidence against the hypothesis, as I
> stated it, is overwhelming.
>
Ohh, but again, you are shifting. At one point we are talking about the odds
of an idependent event (the perfect flip of a coin is 50:50 regardless of
how my times you flip it, and what the outcome was before), then you are
talking about the odds are of getting multiple results (the odds of getting
tails 5 times in a row). They are two completely different things.
> It is more illuminating if the countries are replaced by two mutually
> exclusive universes (you can't be in both at the same time), with
> distinctive fixed rules.
>
> > And remember, it is "A body of evidence" that makes a theory
> > accepted, not a single observation.
>
> You are welcome to modify that thought experiment to be allowed to
> keep sampling people. When another gun owner criminal is sampled, the
> evidence against the hypothesis (gun owner = criminal) will become
> even stonger.
>
> DZ
No, If you have statically predicted number of observations that will be
true, and a number that will be false. The prediction is only strengthened
or weakened when a series of observations creates odds that rises above the
statistics outlined is exceeded.
And this is a problem from a statistician?
--
Dr. Dickie
Skepticult member in good standing #394-00596-438
Hugh Beyer
29726465.180559988.29564.28316.8192:
You're still not clear in what your setup is. Do I know about the US and
UK and the distribution of gun owners in each? Surely not; if I did I
wouldn't have the hypothesis that all gun owners might be criminals. That
being the case I have no why to decide ahead of time what the ratios of
non-gun-owners, law-abiding gun owners, and criminal gun owners might be.
So finding a criminal gun owner can only bolster my hypothesis that all
gun owners are criminals, because it leads me to the prediction that IF I
find a gun owner, he will be a criminal. Prediction met.
The only way you can say that finding reduces the probably of my
hypothesis is by bringing in external information that, by your
assumption, I can't have. Without that information, you just have an
alternative hypothesis, which brings us back to the point that you can't
prove a theory, only disprove it.
DZ
Glad you asked. The way I argued in the UK vs. US problem was along
the lines of a calculation made by a renowned statistician, IJ Good,
when he was considering "Hempel's paradox":
"The hypothesis that all crows are black is the same as that all
non-black things are not crows, and this is supported by the
observation of a white shoe".
(a sub-problem is: how the odds in favor of the hypothesis change once
we observed a black crow?)
>> > In addition, you are dealing with statistics,and the statistics of
>> > small numbers (n=1), as I am sure you very well know, is not even
>> > worth discussing ;-0
>>
>> This is wrong. In statistics, every new observation is shifting the
>> existing odds even if by a small amount.
>
> No, no matter how many times you filp a coin (a perfect coin,
> perfect flip), the odds are still 50:50 you will get tails.
You are confusing the reality i.e. "fair coin" with the statistical
odds that are data dependent. If you would like to make a decision
(based on the data) that the coin is (un)fair, you have to establish a
threshold for the evidence, for example the threshold value of
Pr(fair | data). Now a single piece of evidence can make the ratio go
above the threshold and the resulting inference is equivalent to the
situation where you collected ALL that data at once.
Now the following is proving that assertion, (some philosophers of
science who are preaching us how to do statistics are completely
unaware of) -
Let w0 be the prior probability of the hypothesis H0, that is
Pr(H0). Let p0, q0 be the evidence given H0 and given the alternative
hypothesis (H1) correspondingly. Having seen the first piece of
information (e.g. gun-owner-criminal), the updated Pr(H0 | Data) is
w1 = w0*p0 / (w0*p0 + (1 - w0)*q0)
As another independent piece of evidence arrives, this gives the new
piece of evidence p1 and q1, and the updated Pr(H0 | Data) is
w2 = w1*p1 / (w1*p1 + (1 - w1)*q1)
Here you recalculate (update) the probability given your past
experience.
However, if you do the algebra, this probability (when expressed
through w0) is
w2 = p0*p1*w0 / (p0*p1*w0 + (1 - w0)*q0*q1)
This is the same expression as that for the probability of the
hypothesis after having looked at ALL the data at once, where we have
the usual overall likelihoods as the products of independent pieces,
p0*p1 and q0*q1. This is easily extended to more than two
observations.
>> This is the basis for sequential inference. A decision is made when
>> a single new observation pushes the odds above a certain
>> threshold. In the example considered, any new observation of a
>> criminal gun owner is shifting the odds towards the alternative
>> "not every gun owner is a criminal". If you actually compute the
>> odds, the evidence against the hypothesis, as I stated it, is
>> overwhelming.
>
> Ohh, but again, you are shifting. At one point we are talking about the odds
> of an idependent event (the perfect flip of a coin is 50:50 regardless of
> how my times you flip it, and what the outcome was before), then you are
> talking about the odds are of getting multiple results (the odds of getting
> tails 5 times in a row). They are two completely different things.
See above.
>> It is more illuminating if the countries are replaced by two mutually
>> exclusive universes (you can't be in both at the same time), with
>> distinctive fixed rules.
>>
>> > And remember, it is "A body of evidence" that makes a theory
>> > accepted, not a single observation.
>>
>> You are welcome to modify that thought experiment to be allowed to
>> keep sampling people. When another gun owner criminal is sampled, the
>> evidence against the hypothesis (gun owner = criminal) will become
>> even stonger.
>
> No, If you have statically predicted number of observations that will be
> true, and a number that will be false. The prediction is only strengthened
> or weakened when a series of observations creates odds that rises above the
> statistics outlined is exceeded.
See above. BTW, in the language of "science philosophers", the
prediction vs. accommodation situations are identical, as long as the
accommodating scientist is honest about his hypothesis.
DZ
DZ
>>> Your sample of one is raising the probability of the first and
>>> you've rigged it so that the first implies the second FOR THIS
>>> COUNTRY.
>>
>> That's exactly how I formulated the hypothesis:
>>
>> "...our hypothesis is that all gun owners IN A GIVEN COUNTRY are
>> criminals".
>>
>> Imagine two possible universes "UK" vs. "US" and a mathematical theory
>> that predicts two distinct sets of natural laws for these two
>> worlds. However, only one of the two worlds can be realized, with
>> equal probability. You are to believe the theory. You are in one of
>> the two possible worlds but don't know in which one.
>>
>> In this setup, it doesn't make sense to calculate the probability
>> averaged over two mutually exclusive realizations.
>
> UK and the distribution of gun owners in each? Surely not; if I did I
> wouldn't have the hypothesis that all gun owners might be criminals.
That's not the hypothesis, because in this setup it doesn't make sense
to consider the marginal evidence (i.e. evidence averaged over the
mutually exclusive worlds).
Therefore, the hypothesis is:
"all gun owners IN THE UNIVERSE (whatever one I'm in), are criminals".
> That being the case I have no why to decide ahead of time what the
> ratios of non-gun-owners, law-abiding gun owners, and criminal gun
> owners might be. So finding a criminal gun owner can only bolster
> my hypothesis that all gun owners are criminals, because it leads me
> to the prediction that IF I find a gun owner, he will be a
> criminal. Prediction met.
>
> The only way you can say that finding reduces the probably of my
> hypothesis is by bringing in external information that, by your
> assumption, I can't have.
Yes you can have it. The theory that describes various possible
demographic ratios in those mutually exclusive worlds is axiomatic and
as I said you're welcome to accept it.
> which brings us back to the point that you can't prove a theory,
> only disprove it.
According to Popper, you cannot prove this statement :-)
But in fact I agree. Here is the thing. Consider two men, you and
Popper faced with the same situation.
1) YOU walk along a dusky street and see a figure approaching, still
far away. As you come closer you keep collecting the information
regarding the threat (such as: child vs. adult; male vs. female, armed
vs. unarmed, etc). At some point, the evidence
"likely man; around 35yo; kinda mean-looking; has a gun or something
in the pants"
accumulates to a critical point and prompts you to cross the street to
avoid contact. You can make the decision only because you had a
subjective, prior probability of a threat even before you started
collecting visual information, and you updated it with the
data. Animal and human behavior (correction: this excludes Popper) and
decision making are intrinsically Bayesian. This is completely
equivalent to my offering of the "axiomatic theory" in the US vs. UK,
that you can treat as a prior (yes, it is prior INFORMATION).
2) Popper walks along a dusky street. Axioms are unprovable and priors
are not allowed. Popper gets killed. End of story.
DZ
Seth Breidbart
DZ <31...@2333714574.95544740.21531.350.2491> wrote:
>Glad you asked. The way I argued in the UK vs. US problem was along
>the lines of a calculation made by a renowned statistician, IJ Good,
>when he was considering "Hempel's paradox":
>
>"The hypothesis that all crows are black is the same as that all
>non-black things are not crows, and this is supported by the
>observation of a white shoe".
Consider two hypotheses.
1. All crows are black. (Equivalently, all non-black objects are not
crows.)
2. All crows are white. (Equivalently, all non-white objects are not
crows.)
Assume the facts: crows exist, and non-crows exist.
Now, does a purple cow support either or both of the hypotheses?
Seth
--
Wow! This math stuff works. -- Tom Morley
Seth Breidbart
DZ <21...@1134512194.271934674.175.9328.8526> wrote:
>Seth Breidbart <se...@panix.com> wrote:
>> DZ wrote:
>>>Glad you asked. The way I argued in the UK vs. US problem was along
>>>the lines of a calculation made by a renowned statistician, IJ Good,
>>>when he was considering "Hempel's paradox":
>>>
>>>"The hypothesis that all crows are black is the same as that all
>>>non-black things are not crows, and this is supported by the
>>>observation of a white shoe".
>>
>> Consider two hypotheses.
>>
>> 1. All crows are black. (Equivalently, all non-black objects are not
>> crows.)
>>
>> 2. All crows are white. (Equivalently, all non-white objects are not
>> crows.)
>>
>> Assume the facts: crows exist, and non-crows exist.
>> Now, does a purple cow support either or both of the hypotheses?
>
That's trivial. A purple crow falsifies both hypotheses.
Seth
--
This is mfw, nobody wants to raise the quality of the
discourse. -- Lyle McDonald
Charles
This is really in line with recent research that proves that the time
taken to walk a fortnight is in inverse proportion to the probability
of the number of apples in a bunch of grapes.
HTH
Dr_Dickie
"DZ" <31...@2333714574.95544740.21531.350.2491> wrote in message
news:57...@136316586.146619608.4576.13491.1945...
I understand the probability of the outcome changes with each observation;
however, the odds of a single observation is independent of previous
observations. So, the single observation changes the probability of the next
observation, BUT that does not change the evidence in favor or against the
hypothesis.
Your hypothesis was that certain percentage of gun owners were criminal, and
a certain percentage were not. Therefore, until the N = a level of
significance, there is no support for or against.
You have a non-scientific problem there. In science, the problem should have
an experiment that either invalidates (a negative) or supports (a
non-negative). Since your hypothesis is based on a statistical number of
observations, it cannot invalidate or support until the number of
observations reaches significance.
Also, Popper's assertion that only hypothesis that have a large amount
information are worth considering, is because the more information (the more
things that a hypothesis covers) a hypothesis has, the more easily it is
invalidated (the more it covers, the more likely it is to produce a negative
result when tested); therefore, when it does not produce a negative result,
it is of more interest. I like that!
Dr_Dickie
"DZ" <30...@127934533.314592399.3958.17871.6980> wrote in message
news:22...@92276922.95661623.18640.19140.2236...
> Dr_Dickie <Dr_D...@chembench.com> wrote:
> > DZ wrote:
> >> Dr. Dickie <Dr_D...@chembench.com> wrote:
> >> > "DZ" wrote:
> >> >> Dr_Dickie <Dr_D...@chembench.com> wrote:
> >> > No, a gun owner that is a criminal does not weaken the hypothesis
> >> > that not EVERY gun owner is a criminal, in fact it supports it (it
> >> > is a non-negative result--since not every gun owner is a criminal
> >> > leaves the door open for some gun owners are criminal).
> >>
> >> It does weaken it, if you consider the odds of the hypothesis and its
> >> complement through the formal computation.
> >
> > Weaken is very open term. It does not violate the premise, so how can it
> > weaken it?
>
> You are thinking of the odds as in Pr(heads)/Pr(tails). These have
> nothing to do with the odds of a hypothesis. For our hypothesis H and
> its complement H0 (not all gun owners are criminal), the odds in favor
> of H are Pr(H | data) / Pr(H0 | data), where "|" stands for "given".
>
> When this ratio becomes smaller once a new observation (data) arrives,
> the support for H weakens.
>
> They do exactly this in court with DNA evidence that stands for "data"
> there, and H stands for "you're a perpetrator".
>
> JMW can assure you a single observation (n=1) can put someone into
> jail for a long time :-)
No, the single observation there is include in the database of many, many
other observations that went before it. In DNA analysis you are talking
about considering several (I don't know the exact number) of DNA bands.
Getting exactly the right (for the prosecutor, or wrong if you are the
defendant) number of basepairs for each of the bands considered is like
hitting the numbers of the lotto! (Except in the DNA case the numbers are
reused, so even less likely).
bc
Seth Breidbart wrote:
> In article <57...@136316586.146619608.4576.13491.1945>,
> DZ <31...@2333714574.95544740.21531.350.2491> wrote:
>
> >Glad you asked. The way I argued in the UK vs. US problem was along
> >the lines of a calculation made by a renowned statistician, IJ Good,
> >when he was considering "Hempel's paradox":
> >
> >"The hypothesis that all crows are black is the same as that all
> >non-black things are not crows, and this is supported by the
> >observation of a white shoe".
>
> Consider two hypotheses.
>
> 1. All crows are black. (Equivalently, all non-black objects are not
> crows.)
I joined this late, and didn't read every word, but I have to say that
I don't accept this equivalency anyway. It's considered a logical
equivalency, but only because the word "are" is insinuated as
equivalent to "equals" used in a mathematical sense. This is where
logical induction clashes with mathematics, particularly statistics and
set theory.
If the sentence is re-written, "All black things are crows," and that
postulate were to hold true, then the equivalency applies. Clearly,
this is not the case, and this is because crows do not represent the
entirety of the set of all things in the universe which are black.
Therefore, moving on to all non-black things are not crows does not
apply because the equivalency of "are" and "equals" does not hold in
the first statement.
The modelling being done when scientists and mathemeticians attempt to
fit theories to observation is just that, modelling. Equilibrium
strategies are not absolute, they are predictive. The effort is to
find a calculous which can be used to predict the behavior of a system
by accounting for the previously known behaviors of the system. It is
not the same as the induction conundrum of Hempel's black crows. And,
it is subject to change when further observations contradict it. So,
Einstein's "hmm....that's funny..." From Dr. Dickie's sig, (Did
Einstein really say that?) is the cornerstone of scientific inquiry.
The honest scientist or mathemetician is excited when the observation
is out of character with the hypothesis being tested because it
represents an opportunity for improving the hypothesis.
Now what was it you guys were arguing about again? I lost track.
- bc
DZ
> DZ wrote:
>> Let w0 be the prior probability of the hypothesis H0, that is
>> Pr(H0). Let p0, q0 be the evidence given H0 and given the alternative
>> hypothesis (H1) correspondingly. Having seen the first piece of
>> information (e.g. gun-owner-criminal), the updated Pr(H0 | Data) is
>>
>> w1 = w0*p0 / (w0*p0 + (1 - w0)*q0)
>>
>> As another independent piece of evidence arrives, this gives the new
>> piece of evidence p1 and q1, and the updated Pr(H0 | Data) is
>>
>> w2 = w1*p1 / (w1*p1 + (1 - w1)*q1)
>>
>> Here you recalculate (update) the probability given your past
>> experience.
>>
>> However, if you do the algebra, this probability (when expressed
>> through w0) is
>>
>> w2 = p0*p1*w0 / (p0*p1*w0 + (1 - w0)*q0*q1)
>>
>> This is the same expression as that for the probability of the
>> hypothesis after having looked at ALL the data at once, where we have
>> the usual overall likelihoods as the products of independent pieces,
>> p0*p1 and q0*q1. This is easily extended to more than two
>> observations.
Let me fix your language first. I hope you don't publish with that
mouth :-)
> I understand the probability of the outcome changes with each
> observation;
Not of the "outcome". For example, heads/tails are outcomes in
binomial trials, and the probability of a head doesn't
change. ESTIMATED parameters change, like ESTIMATED Pr(head), and
shockingly, Pr(hypothesis | data) does too.
> however, the odds of a single observation is independent of previous
> observations.
When you talk statistics, reserve "odds" for ratios. PROBABILITY of a
single new observation is independent of previous observations.
> So, the single observation changes the probability of the next
> observation,
Probability of the next observation doesn't change, because as you
just said, our observations are independent.
> BUT that does not change the evidence in favor or against the
> hypothesis.
Yes it does. Having started talking about odds and probabilities, you
can't back off now from formally saying what you mean by the evidence
for a hypothesis. So make your choice now but keep in mind that
magnitudes of all these quantities depend on a SINGLE (new or first)
observation:
1) The evidence in favor of a hypothesis H is the probability of that
hypothesis given the data, Pr(H | data)
2) The evidence in favor of a hypothesis H vs the competing
hypothesis H0 is given by the odds, Pr(H | data)/Pr(H0 | data)
3) The evidence in favor of a hypothesis H vs the competing
hypothesis H0 is given by the factor Pr(data | H)/Pr(data | H0)
> Your hypothesis was that certain percentage of gun owners were
> criminal, and a certain percentage were not. Therefore, until the N
> = a level of significance, there is no support for or against.
Please don't publish with that mouth. The "level of significance" has
nothing to do with support for (or odds against) a hypothesis, because
the attained level of significance is computed as the (cumulative)
probability of the data given the (null) hypothesis ONLY,
i.e. cumulative Pr(data | H0), aka "p-value". Much unlike the
"p-value", Pr(H0 | data) definition (see way above) involves both H
and H0 and therefore properly takes the likelihoods under BOTH
hypotheses into account. Again, keep in mind that Pr(data | H0) used
for significance testing is computed under H0 only, ignoring the
balance between competing hypotheses, thus it cannot be used as a
measure of the plausibility of a hypothesis under any circumstances.
> You have a non-scientific problem there. In science, the problem
> should have an experiment that either invalidates (a negative) or
> supports (a non-negative). Since your hypothesis is based on a
> statistical number of observations, it cannot invalidate or support
> until the number of observations reaches significance.
More alchemistry! Statistical significance has nothing to do with
support of a hypothesis, or with a degree to which the hypothesis is
unlikely.
> Also, Popper's assertion that only hypothesis that have a large
> amount information are worth considering, is because the more
> information (the more things that a hypothesis covers) a hypothesis
> has, the more easily it is invalidated (the more it covers, the more
> likely it is to produce a negative result when tested); therefore,
> when it does not produce a negative result, it is of more
> interest. I like that!
Well, I'm saddened to suggest your excitement is premature.
DZ
Dr. Dickie
"DZ" <20...@268485152.733626583.29378.26502.14542> wrote in message
news:11...@205837038.385013898.23723.32678.26318...
> Dr_Dickie <Dr_D...@chembench.com> wrote:
> > DZ wrote:
> >> Let w0 be the prior probability of the hypothesis H0, that is
> >> Pr(H0). Let p0, q0 be the evidence given H0 and given the alternative
> >> hypothesis (H1) correspondingly. Having seen the first piece of
> >> information (e.g. gun-owner-criminal), the updated Pr(H0 | Data) is
> >>
> >> w1 = w0*p0 / (w0*p0 + (1 - w0)*q0)
> >>
> >> As another independent piece of evidence arrives, this gives the new
> >> piece of evidence p1 and q1, and the updated Pr(H0 | Data) is
> >>
> >> w2 = w1*p1 / (w1*p1 + (1 - w1)*q1)
> >>
> >> Here you recalculate (update) the probability given your past
> >> experience.
> >>
> >> However, if you do the algebra, this probability (when expressed
> >> through w0) is
> >>
> >> w2 = p0*p1*w0 / (p0*p1*w0 + (1 - w0)*q0*q1)
> >>
> >> This is the same expression as that for the probability of the
> >> hypothesis after having looked at ALL the data at once, where we have
> >> the usual overall likelihoods as the products of independent pieces,
> >> p0*p1 and q0*q1. This is easily extended to more than two
> >> observations.
>
> Let me fix your language first. I hope you don't publish with that
> mouth :-)
>
Okay, but again, you are playing a game here. You have a hypothesis based on
a probablity; therefore, until you have sampled the population to signficant
level (I am a chemist here, not a statisician, so forgive the languare-you
know what I mean), you have information, not useful information, just
information. If you are drawing conclusions based on a single observation
for a hypothesis rooted in a probablity of a population, you are kidding
yourself.
> > You have a non-scientific problem there. In science, the problem
> > should have an experiment that either invalidates (a negative) or
> > supports (a non-negative). Since your hypothesis is based on a
> > statistical number of observations, it cannot invalidate or support
> > until the number of observations reaches significance.
>
> More alchemistry! Statistical significance has nothing to do with
> support of a hypothesis, or with a degree to which the hypothesis is
> unlikely.
>
Sure it does. Until you have a proper sample, you cannot get any information
about whether or not a hypothesis is supported or not. Before proper
sampling, you are just making a blind guess!
> > Also, Popper's assertion that only hypothesis that have a large
> > amount information are worth considering, is because the more
> > information (the more things that a hypothesis covers) a hypothesis
> > has, the more easily it is invalidated (the more it covers, the more
> > likely it is to produce a negative result when tested); therefore,
> > when it does not produce a negative result, it is of more
> > interest. I like that!
>
> Well, I'm saddened to suggest your excitement is premature.
>
> DZ
Well try publishing in an analytical journal with an n=1 for your
statistical analysis, and you will be saddened again.
However interesting the problem is in theory, it is ridiculous in real life.
I just cannot see how a single measurement adds to or subtracts from the
confidence of a hypothesis based on a statistical probability.
The experiment must either produce a non-negative or negative answer;
however, because a single measurement does neither, you have nothing.
Know doubt there is some deep level statistician point I am missing;
however, out here in the real world, I say that is crazy.
Not that there is anything wrong with crazy.
DZ
> Okay, but again, you are playing a game here. You have a hypothesis
> based on a probablity;
No, the hypotheses, e.g. H as in "the coin is fair" and H0, "the coin
is not fair" are NOT based on probablity. They're simple statements.
>> Statistical significance has nothing to do with support of a
>> hypothesis, or with a degree to which the hypothesis is unlikely.
>>
> Sure it does.
If you ever practice statistics, this is very important to
understand. The cumulative Pr(data | H0) used for computation of
statistical significance aka "P-value" is computed under H0 only,
ignoring the balance between competing hypotheses. Thus, it cannot be
used as a measure of the (in)plausibility of a hypothesis under any
circumstances.
OTOH, Pr(H0 | data) definition involves likelihoods of the data under
both H and H0 and therefore properly takes BOTH hypotheses into
account.
> Until you have a proper sample, you cannot get any information about
> whether or not a hypothesis is supported or not. Before proper
> sampling, you are just making a blind guess!
Well, you're starting to disagree with Popper here because the notion
of "blind guess" (aka prior) opens a floodgate of plausibility for the
positive evidence.
Next, what sample size (N) is sufficient? Surely, there is no
"universal minimum N" good for every problem. Are you prepared to
argue that N can vary from problem to problem but cannot go as low as
N=1?
If yes, I demand the minimum threshold for the N then :-)
(read: N=1 is enough for certain problems)
> However interesting the problem is in theory, it is ridiculous in
> real life. I just cannot see how a single measurement adds to or
> subtracts from the confidence of a hypothesis based on a statistical
> probability.
You must be trying my patience as I gave you the explanation in
excruciating detail :-) in a post with the notation using w0, w1, w2.
What that showed is that the evidence using ALL the data at once is
equivalent to updating the hypothesis probability by every single new
observation, as it arrives.
DZ
WillBrink
"bc" <wscii_...@comcast.net> wrote:
>
> Now what was it you guys were arguing about again? I lost track.
I was trying to make a point while they were locked in mental
masturbation :-)
>
> - bc
>
--
Will Brink @ http://www.brinkzone.com/
The BillRodgers
"WillBrink" <WillBrink*NOSPAM*@Comcast.net> wrote in message
news:WillBrink*NOSPAM*-88B409.170...@comcast.dca.giganews.com...
> In article <1128009919....@g43g2000cwa.googlegroups.com>,
> "bc" <wscii_...@comcast.net> wrote:
>
>>
>> Now what was it you guys were arguing about again? I lost track.
>
> I was trying to make a point while they were locked in mental
> masturbation :-)
>
DZ
> Seth Breidbart wrote:
>> DZ wrote:
>> >Glad you asked. The way I argued in the UK vs. US problem was along
>> >the lines of a calculation made by a renowned statistician, IJ Good,
>> >when he was considering "Hempel's paradox":
>> >
>> >"The hypothesis that all crows are black is the same as that all
>> >non-black things are not crows, and this is supported by the
>> >observation of a white shoe".
>>
>> Consider two hypotheses.
>>
>> 1. All crows are black. (Equivalently, all non-black objects are not
>> crows.)
>
> I joined this late, and didn't read every word, but I have to say
> that I don't accept this equivalency anyway. It's considered a
> logical equivalency, but only because the word "are" is insinuated
> as equivalent to "equals" used in a mathematical sense. This is
> where logical induction clashes with mathematics, particularly
> statistics and set theory.
It doesn't clash. Both Hempel and Good (mathematical statistician)
understood the problem like this, quoting Good:
The hypothesis H is of the form that class A is contained in class B,
for example "all crows are black".
If you read it differently, then you're arguing about some other
paradox.
DZ
Dr. Dickie
"DZ" <32...@293632974.270899021.27586.30898.32690> wrote in message
news:26...@284955866.1955820155.29838.27260.11426...
> Dr. Dickie <Dr_D...@chembench.com> wrote:
> > Okay, but again, you are playing a game here. You have a hypothesis
> > based on a probablity;
>
> No, the hypotheses, e.g. H as in "the coin is fair" and H0, "the coin
> is not fair" are NOT based on probablity. They're simple statements.
>
> >> Statistical significance has nothing to do with support of a
> >> hypothesis, or with a degree to which the hypothesis is unlikely.
> >>
> > Sure it does.
>
> If you ever practice statistics, this is very important to
> understand. The cumulative Pr(data | H0) used for computation of
> statistical significance aka "P-value" is computed under H0 only,
> ignoring the balance between competing hypotheses. Thus, it cannot be
> used as a measure of the (in)plausibility of a hypothesis under any
> circumstances.
>
> OTOH, Pr(H0 | data) definition involves likelihoods of the data under
> both H and H0 and therefore properly takes BOTH hypotheses into
> account.
>
I understand what you are saying, I understand your premise; however, I just
completely disagee with the premise that each sample builds like bricks to
support or weaken a hypothesis. It is a nice theoretical construct, but
(IMHO) is has no validity in real life. It is like talking about the
accuracy of a single measurement, it simply means nothing.
> > Until you have a proper sample, you cannot get any information about
> > whether or not a hypothesis is supported or not. Before proper
> > sampling, you are just making a blind guess!
>
> Well, you're starting to disagree with Popper here because the notion
> of "blind guess" (aka prior) opens a floodgate of plausibility for the
> positive evidence.
>
No, no, I am not talking about making a blind guess in order to set up an
experiment, I am saying, you are drawing a conclusion before enough data has
been collected. Perhaps I misunderstood the premise of the original problem,
but let me see if I can explain my objection (BTW, I love the dialog--though
we are no doubt boring most of the folks here).
Your hypothesis is 50% of gun owners are criminals. You then take a data
point (a single sample). Regadless of the answer (the sample is a gun owner
and criminal, or not criminal and gun owers) it does not support of weaken
the hypothesis. Yes, it changes the probablity, but it does not call for a
conclusion. Why? If the hypothesis is that 50% of gun owners are criminals
(or 99.999%, it makes no difference), then that allows there to be
non-criminals who are gun owners, so it is within the parameters of the
hypothesis. Only after enough sampling has been done (I know there is a
formula--reaching back to my statisics class from 20 years ago) to calculate
the sample number need before you have enough sampled to reach any
legitimate conclusion.
> Next, what sample size (N) is sufficient? Surely, there is no
> "universal minimum N" good for every problem. Are you prepared to
> argue that N can vary from problem to problem but cannot go as low as
> N=1?
>
> If yes, I demand the minimum threshold for the N then :-)
>
> (read: N=1 is enough for certain problems)
>
> > However interesting the problem is in theory, it is ridiculous in
> > real life. I just cannot see how a single measurement adds to or
> > subtracts from the confidence of a hypothesis based on a statistical
> > probability.
>
> You must be trying my patience as I gave you the explanation in
> excruciating detail :-) in a post with the notation using w0, w1, w2.
> What that showed is that the evidence using ALL the data at once is
> equivalent to updating the hypothesis probability by every single new
> observation, as it arrives.
>
> DZ
Again, I understand what you are saying (I simply have not be able to get my
point across--and used some incorrect langue in my attempts). BUT, the
changes in probability are NOT (again, in my opinion as a practicing
scientist) the same as a change in the outcome. Only after the outcome (that
is, enough sampling has been done to draw conclusions) do the results
support, or weaken the hypothesis. Before that, it is simply raw data being
collected, and drawing conclusion before enough data is irrelevant.
I do see your point, and it is interesting (in the abstract), I shall have
to read some of the papers from the author (and dust off my Popper--as soon
as I finish some current reading).
Have a good weekend (you too Charles).
TheBillRogers
<Dr_D...@chembench.com> wrote:
>I was refering to your hypothesis that 50% of gun owners are criminals.
ALL gun owners are criminals.
TheBillRogers
<33...@108905568.581527700.17030.32179.17379> wrote:
>The hypothesis was that ALL gun owners are criminals,
And they are.
WillBrink
Å‚alternativeË› theory to evolution, will make its way all the way to the
supreme court at some point. ItÄ…s un-fing-believable to me that in the
year 2005, this topic would waste the courts times and these creationist
simpletons would not be ignored at this point. Talk about human
enlightenment being dragged back a few hundred years...thanx GWB for
making the Republican party the political arm of the religious right and
giving power to people that think some imaginary friend created the
universe. Just pitifulÅ .
Court Case Threatens to 'Drag Science into the Supernatural'
Ker Than
LiveScience Staff Writer
LiveScience.com
Thu Sep 22, 9:00 PM ET
A court case that begins Monday in Pennsylvania will be the first to
determine whether it is legal to teach a controversial idea called
intelligent design in public schools.
Intelligent design, often referred to as ID, has been touted in recent
years by a small group of proponents as an alternative to Darwin's
theory of evolution. ID proponents say evolution is flawed. ID asserts
that a supernatural being intervened at some point in the creation of
life on Earth.
Scientists counter that evolution is a well-supported theory and that ID
is not a verifiable theory at all and therefore has no place in a
science curriculum.
The case is called Kitzmiller v. Dover Area School District.
Prominent scientists Thursday called a teleconference with reporters to
say that intelligent design distorts science and would bring religion
into science classrooms.
"The reason this trial is so important is the Dover disclaimer brings
religion straight into science classrooms," said Alan Leshner, the CEO
of the American Association for the Advancement of Science (AAAS) and
executive publisher of the journal Science. "It distorts scientific
standards and teaching objectives established by not only state of
Pennsylvania but also leading scientific organizations of the United
States."
"This will be first legal challenge to intelligent design and we'll see
if they've been able to mask the creationist underpinnings of
intelligent design well enough so that the courts might allow this into
public school," said Eugenie Scott, executive director of the National
Center for Science Education (NCSE), which co-hosted the teleconference.
AAAS is the world's largest general science society and the NCSE is a
nonprofit organization committed to helping ensure that evolution
remains a part of public school curriculums.
The suit was filed by the American Civil Liberties Union (ACLU) on
behalf of concerned parents after Dover school board officials voted 6-3
last October to require that 9th graders be read a short statement about
intelligent design before biology lessons on evolution. Students were
also referred to an intelligent design textbook to learn more
information about the controversial idea.
The Dover school district earlier this month attempted to prevent the
lawsuit from going forward, but a federal judge ruled last week that the
trial would proceed as scheduled.
The lawsuit argues that intelligent design is an inherently religious
argument and a violation of the First Amendment that forbids
state-sponsored schools from funding religious activities.
"Although it may not require a literal reading of Genesis, [ID] is
creationism because it requires that an intelligent designer started or
created and intervened in a natural process," Leshner said. "ID is
trying to drag science into the supernatural and redefine what science
is and isn't."
What the panelists are hoping for is not just decision in favor of the
plaintiffs, but one that is so forceful that the Dover school board will
not risk appealing the case to the Supreme Court and having a negative
ruling with national ramifications.
Scott pointed to an earlier case, McLean v. Arkansas, in which the state
tried to get creation science taught alongside evolution in public
schools.
"What happened was the pounding that creation science got was so solid
that the state didn't even appeal, they just threw in the towel and quit
there," Scott said.
Scott fears that a decision in favor of the Dover school district will
embolden other schools around the country that want to introduce
religious views into their curriculums.
On the other hand, if the plaintiffs win and intelligent design is
declared unconstitutional, then "this will definitely throw sand in the
gears of efforts in other school districts to institute [ID]," said
Scott.
Dr. Dickie
"DZ" <33...@108905568.581527700.17030.32179.17379> wrote in message
news:11...@295367876.7332784.32302.8506.20732...
> >> > Okay, but again, you are playing a game here. You have a hypothesis
> >> > based on a probablity;
> >>
> >> No, the hypotheses, e.g. H as in "the coin is fair" and H0, "the coin
> >> is not fair" are NOT based on probablity. They're simple statements.
> >>
> > I was refering to your hypothesis that 50% of gun owners are criminals.
>
> (or 50%) are criminals :-)
>
> But, the way you said it now, is the same as "the coin is fair" -
> exactly half of gun owners are criminals (no probability involved).
>
> If you'd like to argue a hypothesis formulated as "maybe about 50% of
> gun owners are criminals", go right ahead :-)
>
> Once we observe a criminal gun owner, what changes - the degree of
> maybiness, or the 50% value, or both?
Well 100% is different. But let's use the 50% (or whatever value between 0%
and 100%), this does still involve a probability when sampling.
So, the probability changes with each sample taken; however, what I am
saying is that that change is irrelevant to the hypothesis: that is, it
neither supports nor weakens the hypothesis--which is what I understood you
to be saying.
The reason is, until a level of sampling is done such that the enough of
the population has tested be tested to assure that you have significant
results, the samples cannot be used to make a prediction about the total
population.
Since the level of sampling would be such that further sampling would not
change the percentages (to a significant degree), you finally have something
that is of use in making a prediction (or testing the hypothesis).
The fact that the probability changes with each sampling is interesting in
its implications in theory, but in practicality it does not real have any im
pact.
It seems that I did not really understand the original question, I will get
the paper from the author and give it a read.
Thanks for the thoughts.
dami...@oldman.org
And now ID is coming to NJ
http://tinyurl.com/cqk2z
Knowing that GWB made the Republican party the political arm of the
religious right was one of the main reasons why I voted for Kerry.
And I wouldn't have voted for Kerry against GHW Bush
Hugh Beyer
news:75...@26931964.1023315336.549.28830.22050:
> Good was pondering the black crows in the 60s. Those papers won't
> discuss any of these issues. Good simply said - there are situations
> causing decrease in the odds for the hypothesis like "all crows are
> black" when "another black crow is observed". That's outrageous
> enough.
>
> Your issue is more fundamental since (for reasons I don't know) you
> won't accept the importance of decrease/increase in the odds for a
> hypothesis, with a single observation. The isssue is not stochasticity
> of support brought by a single observation, but the apparent change in
> the odds in the direction opposite to common intuition.
>
> In real life (to which you referred when you said this is all nice but
> "has no validity in real life") the importance of the amount of change
> in the odds is dealt with by incorporating the risk associated with
> making decisions, such as "reject this hypothesis". What is small risk
> for you, might be big risk for someone else.
>
FWIW, DZ, I don't think anyone has really understood the thought
experiment you tried to pose a bunch of posts back. I know I couldn't
figure out what you were trying to set up; Dickie thinks he does but I
think he thinks it's about the probability of the event whereas you want
to talk about the choice between two hypotheses. For this to go anywhere I
think you should try restating the original scenario really cleanly.
Hugh
--
Exercise is a dirty word. Whenever I hear it, I wash my mouth out with
chocolate. ("Ladi")
Lee Michaels
is related to the discussion of this thread. I don't know if anybody else
posted it yet.
http://www.theonion.com/content/node/39512
Dr. Dickie
"Hugh Beyer" <beyer...@acm.org> wrote in message
news:Xns96E3B37CAB417...@130.81.64.196...
Yeah, true that. I have this problem, I cannot let something I do not
understand (or think I do not understand) just slip by and accept it so that
I can get to rest. So, I never really got to read the heart of the problem
because I got hung-up on the changing of probability vs support or
weakening the hypothesis (I guess on that we have to agree to disagree, but
I suspect that has to do with the type of work I do versus other scientific
endeavors).
Since I do not see previous read posts, I will have to dig and get the
original problem back up.
BTW DZ, I did a little searching (I am likely do be working too much this
week to pursue it) and found lots on IJ Good, but nothing that covered
philosophy of science. Did he write a book on the subject?
--
------
Dr. Dickie
"Let be be finale of seem.
The only emperor is the emperor of ice-cream."
-- Wallace Stevens
DZ
> BTW DZ, I did a little searching (I am likely do be working too much
> this week to pursue it) and found lots on IJ Good, but nothing that
> covered philosophy of science. Did he write a book on the subject?
I don't know everything he wrote but what I saw was based on
probability. He said in "When batterer turns murderer" Nature 1995
375(6532):541
"The simple concept of the Bayes factor is basic for legal trials. It
is also basic for medical diagnosis and for the philosophy of science"
and he wrote a lot on that.
Seth Breidbart
Dr_Dickie <Dr_D...@chembench.com> wrote:
>Your hypothesis was that certain percentage of gun owners were criminal, and
>a certain percentage were not. Therefore, until the N = a level of
>significance, there is no support for or against.
Let's try a similar thought experiment, without the religious terms.
There are two urns. One has 2 white balls and 98 clear balls. The
other has 25 white balls, 25 black balls, and 50 clear balls. You
pick one of them, but don't know which. The hypothesis is "half the
colored (non-clear) balls in this urn are black" (equivalently, "this
is urn #2").
To start with, the probability of the hypothesis is 50%. You pull one
ball randomly from the urn. It's white. What is the probability of
the hypothesis now? (It turns out to be 25/27.) That is, pulling a
_white_ ball greatly increases the chances that some of the balls in
the urn are black.
>You have a non-scientific problem there. In science, the problem should have
>an experiment that either invalidates (a negative) or supports (a
>non-negative). Since your hypothesis is based on a statistical number of
>observations, it cannot invalidate or support until the number of
>observations reaches significance.
Nope, every observation (that cannot be predicted with certainty)
changes the probability distribution.
Seth
--
This is mfw, nobody wants to raise the quality of the
discourse. -- Lyle McDonald
DZ
> Dr_Dickie <Dr_D...@chembench.com> wrote:
>> Your hypothesis was that certain percentage of gun owners were
>> criminal, and a certain percentage were not. Therefore, until the N
>> = a level of significance, there is no support for or against.
>
> Let's try a similar thought experiment, without the religious terms.
>
> There are two urns. One has 2 white balls and 98 clear balls. The
> other has 25 white balls, 25 black balls, and 50 clear balls. You
> pick one of them, but don't know which. The hypothesis is "half the
> colored (non-clear) balls in this urn are black" (equivalently, "this
> is urn #2").
>
> To start with, the probability of the hypothesis is 50%. You pull one
> ball randomly from the urn. It's white. What is the probability of
> the hypothesis now? (It turns out to be 25/27.) That is, pulling a
> _white_ ball greatly increases the chances that some of the balls in
> the urn are black.
Thank you! Using my notation from two posts up
(http://groups.google.com/group/misc.fitness.weights/msg/2b18d78bcc4207f9)
w0=1/2; p0=25/100; q0=2/100
and the probability of the hypothesis with this first observation
becomes
w1 = w0*p0 / (w0*p0 + (1 - w0)*q0) = 25/27
We don't even need to worry about what w0 value is to show that w1 is
greater than w0 here, or equivalently that the odds in favor of the
hypothesis ("there are some black balls in the urn we took a ball
from") increase with the observation of a white ball.
Dr_Dickie
"Seth Breidbart" <se...@panix.com> wrote in message
news:di254j$rb2$1...@reader1.panix.com...
I know that, I have never argued that, I understand that. ;-)
I did take statistics back in undergrad (all we did was derive the
statistical models from basic principals). My point is, despite the fact
that the probability changes, it is not valid to draw CONCLUSIONS from an
incomplete sampling. I have never argued about the changes in probability, I
argue about drawing conclusions (scientific conclusions) from under
sampling.
If I am trying to determine the percentage of gold in ore, and the ore is
coming in by the train car load. Would you trust an assay of one rock out of
one car? If so, don't buy gold ore.
A single sample from the urn changes the probability, but it is insufficient
to draw conclusions from (you cannot say that this single sample supports or
weakens your hypothesis). Precisely because of the problem that you
outlined, you cannot even tell which urn you are testing. Further sampling
(unless from the correct urn) would show that your hypothesis has problems.
Dr_Dickie
"DZ" <12...@236879751.719622035.19672.26537.19285> wrote in message
news:24...@1607211385.1291219779.3848.18782.22937...
I guess I must be speaking a different language. I understand that (I really
do, I am not just saying that--I am sure that I do not understand it to the
level of a statistician, but I do understand that the probability changes).
I understand that the odds increase in favor of the hypothesis, my point has
always been about drawing conclusions before sufficient testing.
Let me put it this way: Once 85 of the balls have been tested, will the
probability continue to change significantly with each draw?
Dr_Dickie
"Seth Breidbart" <se...@panix.com> wrote in message
news:di254j$rb2$1...@reader1.panix.com...
Right, you are making my point. The probability changes significantly with
each early draw, therefore, you cannot draw conclusions until the population
has been sampled sufficiently. Will the probability distribution continue to
change significantly when the 85th ball is observed?
Hugh Beyer
news:di254j$rb2$1...@reader1.panix.com:
> In article <1127992641.0bba9b772c5bbe00e4fb26b4eddb221d@teranews>,
> Dr_Dickie <Dr_D...@chembench.com> wrote:
>
>>Your hypothesis was that certain percentage of gun owners were criminal,
>>and a certain percentage were not. Therefore, until the N = a level of
>>significance, there is no support for or against.
>
> Let's try a similar thought experiment, without the religious terms.
>
> There are two urns. One has 2 white balls and 98 clear balls. The
> other has 25 white balls, 25 black balls, and 50 clear balls. You
> pick one of them, but don't know which. The hypothesis is "half the
> colored (non-clear) balls in this urn are black" (equivalently, "this
> is urn #2").
>
> To start with, the probability of the hypothesis is 50%. You pull one
> ball randomly from the urn. It's white. What is the probability of
> the hypothesis now? (It turns out to be 25/27.) That is, pulling a
> _white_ ball greatly increases the chances that some of the balls in
> the urn are black.
>
Thank you for finally stating clearly the proposition DZ was trying to put
forward.
My point on this is that DZ's original statement is incorrect: "Let me
assure you that a result in agreement with a hypothesis (i.e. a
confirmatory, or positive result) in fact DECREASES the evidence for that
hypothesis!"
He could only make the statement through a trick. When the 2 possibilities
are known in advance (urn #1 and #2) then yes, finding a white ball
increases the probability that there are black balls in the urn. But your
hypothesis "this is urn #2" or "there are lots of black balls in this
urn" is not equivalent to "I am more likely to pull a white ball out of
urn #1 than #2" and does not lead to the prediction "I will not pull out a
white ball on my first trial."
The only way those predictions make sense is if you *already know* the
distribution of balls in urns 1 & 2.
Hugh Beyer
news:1128599613.0c47590aa56cf83625c8dc0aa64d2021@teranews:
You're trying to make this a sampling problem and it's really not. The
probability of drawing a white ball from urn 1 is 1 out of 49. The
probability of drawing a white ball from urn 2 is 1 out 4. Clearly, if
you're betting, having seen the white ball you would do better to put your
money on its being urn #2--but only because you know the distribution of
balls in the urns.
DZ
> se...@panix.com (Seth Breidbart) wrote
>> Dr_Dickie <Dr_D...@chembench.com> wrote:
>>>Your hypothesis was that certain percentage of gun owners were criminal,
>>>and a certain percentage were not. Therefore, until the N = a level of
>>>significance, there is no support for or against.
>>
>> Let's try a similar thought experiment, without the religious terms.
>>
>> There are two urns. One has 2 white balls and 98 clear balls. The
>> other has 25 white balls, 25 black balls, and 50 clear balls. You
>> pick one of them, but don't know which. The hypothesis is "half the
>> colored (non-clear) balls in this urn are black" (equivalently, "this
>> is urn #2").
>>
>> To start with, the probability of the hypothesis is 50%. You pull one
>> ball randomly from the urn. It's white. What is the probability of
>> the hypothesis now? (It turns out to be 25/27.) That is, pulling a
>> _white_ ball greatly increases the chances that some of the balls in
>> the urn are black.
>
> Thank you for finally stating clearly the proposition DZ was trying to put
> forward.
>
> My point on this is that DZ's original statement is incorrect: "Let me
> assure you that a result in agreement with a hypothesis (i.e. a
> confirmatory, or positive result) in fact DECREASES the evidence for that
> hypothesis!"
>
> He could only make the statement through a trick.
Well you can thank Seth all you want but if I had described the
problem his way we wouldn't have had this lovely conversation :-)
But I take issue with your saying there is a trick or that my
statement was incorrect.
My hypothesis was (in the language of this new formulation)
"In the urn - whatever one we are drawing from - all colored balls
are white".
I claimed that the observation of a white ball undermines this
hypothesis.
> The only way those predictions make sense is if you *already know*
> the distribution of balls in urns 1 & 2.
Knowledge of the distribution doesn't have to be sharp like in this
example. There might be a theory that predicts two or more types of
urns with frequencies of balls in each type characterized only to a
degree. Urns can be popping up according to a probability
distribution.
I would argue that all rational decisions are based on (sometimes
subconscious) calculation similar to the above.
Hugh Beyer
news:16...@358616585.495814498.21777.11095.10772:
I don't argue that rational decisions are made on incremental data, which
seems to be your real point. But the above example doesn't show what you
want to show.
You claim that drawing a white ball is in agreement with your hypothesis
that all the colored balls in the urn are white. It's not. It's better
evidence of the opposite hypothesis--because you know the percentage of
white balls is higher in the urn that also has black balls.
Seth Breidbart
Dr_Dickie <Dr_D...@chembench.com> wrote:
>I guess I must be speaking a different language. I understand that (I really
>do, I am not just saying that--I am sure that I do not understand it to the
>level of a statistician, but I do understand that the probability changes).
>I understand that the odds increase in favor of the hypothesis, my point has
>always been about drawing conclusions before sufficient testing.
>Let me put it this way: Once 85 of the balls have been tested, will the
>probability continue to change significantly with each draw?
Here's the issue: In the examples (specifically the urn one), there
are only two cases. You picked one urn or the other one. In that
example, after a lot fewer than 85 balls, you _know_ which urn you
have.
In the real world, there are a lot more than two cases. The urn has
some number of balls of each type, independently chosen from a
probability distribution. The issue is estimating that distribution
by sampling from the urn. (That is, to fill the urn pick a clear ball
with probability P, a white ball with probability Q, and a black ball
with probability R; P+Q+R=1. Now by sampling, estimate P, Q, and R.)
In that case, you generally gain some information with each sample,
but the amount of information decreases with each subsequent sample.
Seth
--
Wow! This math stuff works. -- Tom Morley
Dr. Dickie
"DZ" <8904@51888066.2748317576.3222.5437.26191> wrote in message
news:31...@303429701.43215320.20167.30451.29843...
> Dr_Dickie <Dr_D...@chembench.com> wrote:
> single draw I would make you broke :-)
Which is why I would never draw conclusions from a single draw ;-)
Dr. Dickie
but I get that); however, you do not get enough information from the first
draw to make conclusions, despite that fact that the information diminishes
with increased draws. At first each draw increases your knowledge
significantly, after a certain number of draws (DZ jogged my memory with
power analysis), additional draws will yield little real information.
Dr. Dickie
you are drawing balls from; therefore, the amount of sampling becomes
important.
Even if you only have one urn, drawing a conclusion (and saying that it
supports or weakens your hypothesis is drawing a conclusion) without enough
information is not correct.
Dr. Dickie
Which gets very much to what I have been saying. Yes, a single study is used
to lead to further investigations; however, it is the body of evidence that
wins the argument in the end. One study is interesting, fifty studies is on
to something, one hundred studies and you can start working on a law or
theory (Obviously these numbers are drawn from the urn of air ;-)
Anywhoo, I am about to be swamped for a while.
Have a good weekend one and all (yes, Charles, that means you!).
Hugh Beyer
news:1128681272.2e0abd40699fc844f8d546d1e36c4a74@teranews:
Pooh. Observing the sun's gravity warping light was a single observation;
but it was powerful evidence in favor of relativity.
Seth Breidbart
Dr. Dickie <Dr_D...@chembench.com> wrote:
>"DZ" <8904@51888066.2748317576.3222.5437.26191> wrote in message
>news:31...@303429701.43215320.20167.30451.29843...
>> Let me put it this way - if we were to bet every evening based on a
>> single draw I would make you broke :-)
>
>Which is why I would never draw conclusions from a single draw ;-)
And if you bet against me, and I have the information from a single
draw (and you don't), I still get to make you broke.
Seth Breidbart
Dr. Dickie <Dr_D...@chembench.com> wrote:
>No problem, I understand that (although it is a bit more complex than that,
>but I get that); however, you do not get enough information from the first
>draw to make conclusions, despite that fact that the information diminishes
>with increased draws.
Define "conclusions". You never know the exact probabilities; all you
can do is give estimates of their bounds (e.g. "With 95% probability,
P is between 0.3 and 0.5")
Andrzej Rosa
>> But in science, that is where this started, the problem is knowing which
>> urn you are drawing balls from; therefore, the amount of sampling
>> becomes important.
>> Even if you only have one urn, drawing a conclusion (and saying that it
>> supports or weakens your hypothesis is drawing a conclusion) without
>> enough information is not correct.
>
> Pooh. Observing the sun's gravity warping light was a single observation;
> but it was powerful evidence in favor of relativity.
>
> Hugh
You know full well, I guess, that Eddington's experiment was not
conducted properly. It "worked" because they knew what they should get.
Bad argument. ;)
--
Andrzej Rosa 1127R
Hugh Beyer
816529063.2201415687.17534.28246.11506:
> Dr. Dickie <Dr_D...@chembench.com> wrote:
>> "Hugh Beyer" <beyer...@acm.org> wrote:
>>> DZ wrote:
>>>> My hypothesis was (in the language of this new formulation)
>>>> "In the urn - whatever one we are drawing from - all colored balls
>>>> are white".
>>>
>>> You claim that drawing a white ball is in agreement with your hypothesis
>>> that all the colored balls in the urn are white. It's not. It's better
>>> evidence of the opposite hypothesis--because you know the percentage of
>>> white balls is higher in the urn that also has black balls.
>>
>> Which gets very much to what I have been saying. Yes, a single study
>> is used to lead to further investigations; however, it is the body
>> of evidence that wins the argument in the end.
>
> No it doesn't. Hugh is arguing that my original statement was
> misleading. That I agree with, but I don't think he expressed any
> objection to the way I re-stated the problem:
>
> "In the urn - whatever one we are drawing from - all colored balls are
> white. The observation of a white ball undermines this hypothesis".
It wasn't misleading, it was confused. It's hard to refute a statement when
you can't be sure what it says.
I don't object to your re-statement, but the observation of a white ball on
a single draw in this problem does NOT agree with the hypothesis that all
colored balls in the urn are white. So your original statement wasn't just
misleading, it was wrong.
DZ
> DZ wrote:
>> "In the urn - whatever one we are drawing from - all colored balls
>> are white. The observation of a white ball undermines this
>> hypothesis".
>
> It wasn't misleading, it was confused. It's hard to refute a
> statement when you can't be sure what it says. I don't object to
> your re-statement, but the observation of a white ball on a single
> draw in this problem does NOT agree with the hypothesis that all
> colored balls in the urn are white. So your original statement
> wasn't just misleading, it was wrong.
Ok, that's good enough for me. I'm content with saying it your way.
So let's re-phrase it back now in terms of gun owners and criminals:
"Our hypothesis is: In this country - whatever one we're in now - all
gun owners are criminals. Observation of a criminal gun owner
undermines this hypothesis".
Given this formulation, would you retract the following:
> So finding a criminal gun owner can only bolster my hypothesis that
> all gun owners are criminals, because it leads me to the prediction
> that IF I find a gun owner, he will be a criminal. Prediction met.
Hugh Beyer
2614817785.973726725.21637.14346.15063:
Yes, because when I wrote that I couldn't tell whether your observer *knew*
that there were only two countries with exactly the distribution of gun
owners, non-owners, and criminals you stated. If that's just one hypothesis
out of many, my statement stands; if the only question is which of the two
predefined countries the observer is in, the observation of any gun owner at
all in England is so unlikely that observing one, criminal or not, supports
the hypothesis that you are in America.
Seth Breidbart
Hugh Beyer <beyer...@acm.org> wrote:
>DZ <23...@1443027177.2550128372.23188.22070.32498> wrote in news:14393@
>816529063.2201415687.17534.28246.11506:
>> Dr. Dickie <Dr_D...@chembench.com> wrote:
>>> "Hugh Beyer" <beyer...@acm.org> wrote:
>>>> DZ wrote:
>>>>> My hypothesis was (in the language of this new formulation)
>>>>> "In the urn - whatever one we are drawing from - all colored balls
>>>>> are white".
>>>>
>>>> You claim that drawing a white ball is in agreement with your hypothesis
>>>> that all the colored balls in the urn are white. It's not.
No, it _is_ in agreement with that hypothesis: it doesn't falsify it.
>>>> It's better
>>>> evidence of the opposite hypothesis--because you know the percentage of
>>>> white balls is higher in the urn that also has black balls.
That's also true. It _agrees with_ both the hypothesis and its
negation; but it increases the probability of the hypothesis's
negation.
>> No it doesn't. Hugh is arguing that my original statement was
>> misleading. That I agree with, but I don't think he expressed any
>> objection to the way I re-stated the problem:
>>
>> "In the urn - whatever one we are drawing from - all colored balls are
>> white. The observation of a white ball undermines this hypothesis".
>
>It wasn't misleading, it was confused. It's hard to refute a statement when
>you can't be sure what it says.
The use of "undermines" is (I think) wrong; the word is too strong for
an event that merely lowers the probability.
>I don't object to your re-statement, but the observation of a white ball on
>a single draw in this problem does NOT agree with the hypothesis that all
>colored balls in the urn are white.
It certainly does agree with it. Observing a non-white colored ball
disagrees with it.
Dr. Dickie
"DZ" <23...@1443027177.2550128372.23188.22070.32498> wrote in message
news:14...@816529063.2201415687.17534.28246.11506...
> >>> "In the urn - whatever one we are drawing from - all colored balls
> >>> are white".
> >>
hypothesis
> >> that all the colored balls in the urn are white. It's not. It's better
> >> evidence of the opposite hypothesis--because you know the percentage of
> >> white balls is higher in the urn that also has black balls.
> >
> > is used to lead to further investigations; however, it is the body
> > of evidence that wins the argument in the end.
>
> misleading. That I agree with, but I don't think he expressed any
> objection to the way I re-stated the problem:
>
But you are basing the "undermining of the hypothesis" on the fact that the
pulling of the white ball reduces the probability of pulling a white ball on
the next draw? Is that correct?
Dr. Dickie
Right, observation of a white ball on the first draw supports the
hypothesis. It reduces the probability of another draw producing a white
ball, but that does not weaken the hypothesis because support or weakening
is based on the evidence, and the evidence is based on what has been
observed (you cannot use a priori information except for the purpose of
intellectual discussion--which we are here).
It does not weaken the hypothesis, it simply increases the probability of
determining that the hypothesis if false (which is the whole point for
sampling in the first place!). This is not a bad thing, it is a good thing
and it is the whole point for looking in the first place. You cannot view
each sample as weakening, since the hypothesis is an either or (100%), it is
either supported (not falsified) or falsified.
Hugh Beyer
> "Seth Breidbart" <se...@panix.com> wrote in message
> news:dicq6o$ph5$1...@reader1.panix.com...
>> In article <Xns96E9E001FFC5C...@130.81.64.196>,
>> Hugh Beyer <beyer...@acm.org> wrote:
>> >DZ <23...@1443027177.2550128372.23188.22070.32498> wrote in
>> >news:14393@ 816529063.2201415687.17534.28246.11506:
>> >> Dr. Dickie <Dr_D...@chembench.com> wrote:
>> >>> "Hugh Beyer" <beyer...@acm.org> wrote:
>> >>>> DZ wrote:
>> >>>>> My hypothesis was (in the language of this new formulation)
>> >>>>> "In the urn - whatever one we are drawing from - all colored
>> >>>>> balls are white".
>> >>>>
>> >>>> You claim that drawing a white ball is in agreement with your
>> >>>> hypothesis that all the colored balls in the urn are white. It's
>> >>>> not.
>>
>> No, it _is_ in agreement with that hypothesis: it doesn't falsify it.
>>
>> >>>> It's better
>> >>>> evidence of the opposite hypothesis--because you know the
>> >>>> percentage of white balls is higher in the urn that also has black
>> >>>> balls.
>>
>> That's also true. It _agrees with_ both the hypothesis and its
>> negation; but it increases the probability of the hypothesis's
>> negation.
You're using different words to say what I said: it's better evidence in
favor of the other hypothesis. You wouldn't expect to see a white ball given
your hypothesis. It's true that it doesn't falsify the hypothesis in the way
that a black ball would, but it's not evidence in support of the hypothesis.
OK, I thought DZ's thing was just stupid, but since you're both falling for
it, I guess not. (There is an alternative hypothesis that I won't go into.)
You're falling for DZ's statement of the hypothesis: "All colored balls in
this urn are white." But because you know the contents of the two urns,
that's really the same as saying: "98% of the balls in this urn are not
white." When the hypothesis is stated that way, do you think drawing a white
ball supports it? Of course not.
Hugh Beyer
No, it's got nothing to do with what happens on the next draw.
DZ
> DZ wrote:
>> "In the urn - whatever one we are drawing from - all colored balls are
>> white. The observation of a white ball undermines this hypothesis".
>
> But you are basing the "undermining of the hypothesis" on the fact
> that the pulling of the white ball reduces the probability of
> pulling a white ball on the next draw? Is that correct?
Not really. For the first draw it doesn't matter whether the urns are
so large that the proportions don't change with the drawings. The case
of "very large" urns is equivalent to "sampling with replacement". If
you're in the middle of having sampled about a half of the urn and
still not sure, there are sampling paths where unusual sampling would
lead to different tentative conclusions (in "with replacement"
vs. "without replacement" scenarios).
As an aside note, "Observation of a purple cow undermines the
hypothesis that all crows are black" :-)
(purple cows in this problem are clear balls, black crows are white
balls, and colored balls are crows; black balls here are unusual so
"black ball" stands for "white crow")
Here is a computer experiment where the first draw happens to be a
white ball and Urn #2 is chosen in reality (which we determine with
certainty at Draw #5).
Urn1: { W=2, B=0, C=98 } (i.e. numbers of "white", black", "clear" balls)
Urn2: { W=25, B=25, C=50 }
(1) Sampling without replacement.
Urn chosen: #2
Draw | { W B C } | Prob(Urn #1) | current sample (W,B,C)
-------------------------------------------------
1 | { 1 0 0 } | 0.07407407 | 1 0 0
2 | { 0 0 1 } | 0.13554633 | 1 0 1
3 | { 1 0 0 } | 0.01238860 | 2 0 1
4 | { 0 0 1 } | 0.02399626 | 2 0 2
5 | { 0 1 0 } | 0.00000000 | 2 1 2
(2) Sampling with replacement
(i.e. "very large" urns with the same proportions)
Urn chosen: #2
Draw | { W B C } | Prob(Urn #1) | current sample (W,B,C)
-------------------------------------------------
1 | { 1 0 0 } | 0.07407407 | 1 0 0
2 | { 0 0 1 } | 0.13554633 | 1 0 1
3 | { 1 0 0 } | 0.00649093 | 2 0 1
4 | { 0 0 1 } | 0.01276820 | 2 0 2
5 | { 0 1 0 } | 0.00000000 | 2 1 2
Now turn one clear ball in Urn1 into a black ball and start over.
Urn1: { W=2, B=1, C=97 }
Urn2: { W=25, B=25, C=50 }
(1) Sampling without replacement.
Urn chosen: #2
Draw | { W B C } | Prob(Urn #1) | current sample (W,B,C)
-------------------------------------------------
1 | { 1 0 0 } | 0.07407407 | 1 0 0
2 | { 0 0 1 } | 0.13434903 | 1 0 1
3 | { 1 0 0 } | 0.00642512 | 2 0 1
4 | { 0 0 1 } | 0.01251088 | 2 0 2
5 | { 0 1 0 } | 0.00050652 | 2 1 2
6 | { 0 0 1 } | 0.00100199 | 2 1 3
7 | { 0 0 1 } | 0.00200197 | 2 1 4
8 | { 0 1 0 } | 0.00000000 | 2 2 4
(2) Sampling with replacement.
Urn chosen: #2
Draw | { W B C } | Prob(Urn #1) | current sample (W,B,C)
-------------------------------------------------
1 | { 1 0 0 } | 0.07407407 | 1 0 0
2 | { 0 0 1 } | 0.13434903 | 1 0 1
3 | { 1 0 0 } | 0.01226373 | 2 0 1
4 | { 0 0 1 } | 0.02352050 | 2 0 2
5 | { 0 1 0 } | 0.00096255 | 2 1 2
6 | { 0 0 1 } | 0.00186567 | 2 1 3
7 | { 0 0 1 } | 0.00361306 | 2 1 4
8 | { 0 1 0 } | 0.00014503 | 2 2 4
9 | { 0 1 0 } | 0.00000580 | 2 3 4
10 | { 0 1 0 } | 0.00000023 | 2 4 4
11 | { 0 0 1 } | 0.00000045 | 2 4 5
12 | { 0 0 1 } | 0.00000087 | 2 4 6
13 | { 0 1 0 } | 0.00000003 | 2 5 6
14 | { 0 1 0 } | 0.00000000 | 2 6 6
DZ
> hypothesis. It reduces the probability of another draw producing a white
> ball,
This is irrelevant, see my other post on sampling with vs. without
replacement.
> but that does not weaken the hypothesis because support or weakening
> is based on the evidence, and the evidence is based on what has been
> observed (you cannot use a priori information except for the purpose of
> intellectual discussion--which we are here).
You're making two mistakes here
1) The evidence here is contained in having drawn a white ball.
2) ALL rational inference including that in statistics (and more
generally, in science) incorporates a priory information (or guesses
for likelihoods for various states of reality, or prior "beliefs").
Consider a simple problem of determining that BMI of a population is
between some numbers X and Y with 95% confidence. Can you take a
sample, compute "95% confidence interval" and make such a statement?
The answer is NO. It's because confidence intervals are random
quantities covering a fixed population value, whereas in the problem
as stated above the population parameter is random, falling in the
fixed interval [X, Y] with 95% probability. You can only make such a
statement if you specify some prior vague (aka "flat") beliefs on what
the distribution of BMI can be.
When you compute dreaded "significance" (or a p-value) concerning a
hypothesis, you subconsciously, yet let me add GENEROUSLY, allow for
the alternative hypothesis to be true. That is an implicit prior
knowledge specification. Otherwise the only (based solely on the
evidence) conclusion you can make is: "a very unlikely event has
occurred under the hypothesis that I know is true". Because it is the
only hypothesis you had allowed. This is a statement that cannot be
used to make decisions regarding choice between hypotheses.
DZ
> Dr. Dickie <Dr_D...@chembench.com> wrote:
>> Right, observation of a white ball on the first draw supports the
>> hypothesis. It reduces the probability of another draw producing a white
>> ball,
>
> This is irrelevant, see my other post on sampling with vs. without
> replacement.
>
>> but that does not weaken the hypothesis because support or weakening
>> is based on the evidence, and the evidence is based on what has been
>> observed (you cannot use a priori information except for the purpose of
>> intellectual discussion--which we are here).
>
> You're making two mistakes here
>
> 1) The evidence here is contained in having drawn a white ball.
>
> 2) ALL rational inference including that in statistics (and more
> generally, in science) incorporates a priory information (or guesses
> for likelihoods for various states of reality, or prior "beliefs").
>
> Consider a simple problem of determining that BMI of a population
Just for clarity, read "mean BMI of a population", i.e. it's a single
number.
Dr. Dickie
"DZ" <31656@3083313108.139985852.13231.1909.30763> wrote in message
news:61...@885915120.591020721.29045.15050.30217...
That was what I said.
> 2) ALL rational inference including that in statistics (and more
> generally, in science) incorporates a priory information (or guesses
> for likelihoods for various states of reality, or prior "beliefs").
>
Well, I kinda disagree with that. Yes prior knowledge is need to make a
hypothesis; however, prior knowledge of the complete system (as in this
case) is not.
> Consider a simple problem of determining that BMI of a population is
> between some numbers X and Y with 95% confidence. Can you take a
> sample, compute "95% confidence interval" and make such a statement?
>
> The answer is NO. It's because confidence intervals are random
> quantities covering a fixed population value, whereas in the problem
> as stated above the population parameter is random, falling in the
> fixed interval [X, Y] with 95% probability. You can only make such a
> statement if you specify some prior vague (aka "flat") beliefs on what
> the distribution of BMI can be.
>
> When you compute dreaded "significance" (or a p-value) concerning a
> hypothesis, you subconsciously, yet let me add GENEROUSLY, allow for
> the alternative hypothesis to be true. That is an implicit prior
> knowledge specification. Otherwise the only (based solely on the
> evidence) conclusion you can make is: "a very unlikely event has
> occurred under the hypothesis that I know is true". Because it is the
> only hypothesis you had allowed. This is a statement that cannot be
> used to make decisions regarding choice between hypotheses.
Of course, that is what the confidence interval says. It is unlikely that
the other hypothesis is true; however, it is always possible. As is any
hypothesis (or theory for that matter), at any time evidence can come along
which negates it, and it is modified (or overhauled) to fit within the new
evidence.