Religion VS Science

It appears that whether it¹s constitutional to teach ID in schools as an
³alternative² theory to evolution, will make its way all the way to the
supreme court at some point. It¹s 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 pitifulŠ.

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.

--
Will Brink @ http://www.brinkzone.com/

Dr. Dickie said:

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.

Spoke like a typical pain-in-the-butt reviewer after having had a
glance over a mathematical proof :-)

Quoted message said:

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.

That formula (used for power calculations) is useless!

(I published a couple of those myself).

It gives the sample size N needed for the statistic V that follows the
distribution W, to exceed certain value X with probability Y, assuming
population parameters Z.

All these things affect N but what is more important, this formula is
barking up the wrong tree, because once again it only considers ONE
aka "alternative" hypothesis, determined by fixing parameter values in
Z.

"DZ" <[email hidden]> wrote in message
news:[email hidden]...

Quoted message said:

Dr. Dickie said:

DZ said:

Dr. Dickie <[email hidden]> 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.

I was refering to your hypothesis that 50% of gun owners are criminals.

The hypothesis was that ALL gun owners are criminals, not MAYBE 100%
(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.

--
Dr. Dickie
Skepticult member in good standing #394-00596-438
Poking kooks with a pointy stick

WillBrink said:

It appears that whether it¹s constitutional to teach ID in schools as an
³alternative² theory to evolution, will make its way all the way to the
supreme court at some point. It¹s 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 pitifulŠ.

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.

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

Dr. Dickie said:

DZ said:

Dr. Dickie said:

DZ wrote:
> Dr. Dickie <[email hidden]> 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.

I was refering to your hypothesis that 50% of gun owners are criminals.

The hypothesis was that ALL gun owners are criminals, not MAYBE 100%
(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.

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.

DZ <[email hidden]> wrote in
news:[email hidden]:

Quoted message said:

Dr. Dickie said:

DZ said:

Dr. Dickie <[email hidden]> wrote:
> DZ wrote:
>> Dr. Dickie <[email hidden]> 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.
>
> I was refering to your hypothesis that 50% of gun owners are
> criminals.

The hypothesis was that ALL gun owners are criminals, not MAYBE 100%
(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.

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"😉

I haven't been following this thread. I saw this on another newsgroup. It
is related to the discussion of this thread. I don't know if anybody else
posted it yet.

http://www.venganza.org/

http://www.theonion.com/content/node/39512

"Hugh Beyer" <[email hidden]> wrote in message
news:[email hidden]...

Quoted message said:

DZ <[email hidden]> wrote in
news:[email hidden]:

Quoted message said:

Dr. Dickie said:

DZ wrote:
> Dr. Dickie <[email hidden]> wrote:
> > DZ wrote:
> >> Dr. Dickie <[email hidden]> 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.
> >
> > I was refering to your hypothesis that 50% of gun owners are
> > criminals.
>
> The hypothesis was that ALL gun owners are criminals, not MAYBE 100%
> (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.

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

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

Dr. Dickie said:

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.

In article <1127992641.0bba9b772c5bbe00e4fb26b4eddb221d@teranews>,

Dr_Dickie said:

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.

Quoted message said:

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

Seth Breidbart said:

Dr_Dickie said:

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.

"Seth Breidbart" <[email hidden]> wrote in message
news:[email hidden]...

Quoted message said:

In article <1127992641.0bba9b772c5bbe00e4fb26b4eddb221d@teranews>,

Dr_Dickie said:

Your hypothesis was that certain percentage of gun owners were criminal,

and

Quoted message said:

Quoted message said:

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.

Quoted message said:

You have a non-scientific problem there. In science, the problem should

have

Quoted message said:

Quoted message said:

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

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
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" <[email hidden]> wrote in message
news:[email hidden]...

Quoted message said:

Seth Breidbart said:

Dr_Dickie said:

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.

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
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

"Seth Breidbart" <[email hidden]> wrote in message
news:[email hidden]...

Quoted message said:

In article <1127992641.0bba9b772c5bbe00e4fb26b4eddb221d@teranews>,

Dr_Dickie said:

Your hypothesis was that certain percentage of gun owners were criminal,

and

Quoted message said:

Quoted message said:

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.

Quoted message said:

You have a non-scientific problem there. In science, the problem should

have

Quoted message said:

Quoted message said:

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
--

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?
--
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

[email hidden] (Seth Breidbart) wrote in
news:[email hidden]:

Quoted message said:

In article <1127992641.0bba9b772c5bbe00e4fb26b4eddb221d@teranews>,

Dr_Dickie said:

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

--
Exercise is a dirty word. Whenever I hear it, I wash my mouth out with
chocolate. ("Ladi"😉

"Dr_Dickie" <[email hidden]> wrote in
news:1128599613.0c47590aa56cf83625c8dc0aa64d2021@teranews:

Quoted message said:

"Seth Breidbart" <[email hidden]> wrote in message
news:[email hidden]...

Quoted message said:

In article <1127992641.0bba9b772c5bbe00e4fb26b4eddb221d@teranews>,

Dr_Dickie said:

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.

Quoted message said:

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

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.

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.

Hugh

--
Exercise is a dirty word. Whenever I hear it, I wash my mouth out with
chocolate. ("Ladi"😉

Dr_Dickie said:

DZ said:

Seth Breidbart said:

Dr_Dickie <[email hidden]> 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.

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?

Let me put it this way - if we were to bet every evening based on a
single draw I would make you broke :-)

Hugh Beyer said:

[email hidden] (Seth Breidbart) wrote

Quoted message said:

Dr_Dickie said:

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.

Quoted message said:

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.

DZ <[email hidden]> wrote in
news:[email hidden]:

Quoted message said:

Hugh Beyer said:

[email hidden] (Seth Breidbart) wrote

Quoted message said:

Dr_Dickie <[email hidden]> 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.

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.

Hugh

--
Exercise is a dirty word. Whenever I hear it, I wash my mouth out with
chocolate. ("Ladi"😉

In article <1128599999.fb68e2e5c7738ce02493d82d37fca31b@teranews>,

Dr_Dickie said:

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