The Impossible

[super animator]

I don't know about you guys, but this debate is sure is interesting.

 

[super animator]

Also to your op, it's impossible for a theory to be completely unfalsifiable.

 

[Upisoft]

Also to your op, it's impossible for a theory to be completely unfalsifiable.

OK, I'll try. There are exactly 5000 universes, but there is no way to travel between them.
Falsify this.

 

[super animator]

OK, I'll try. There are exactly 5000 universes, but there is no way to travel between them.
Falsify this.

That's more of a hypothesis than a theory.

 

[CabVet]

That's more of a hypothesis than a theory.

I wouldn't call it a hypothesis either, unless who proposed it can come up with some sort of rationale.

 

[super animator]

I wouldn't call it a hypothesis either, unless who proposed it can come up with some sort of rationale.

Yea, I just realized that, it's more of a guess or conjure than anything.

 

[Davian]

I don't know about you guys, but this debate is sure is interesting.

Indeed, but I was getting dizzy from watching Resha go around in circles.

 

[Resha Caner]

And another refusal to supply specifics. Go figure.

I offered the data for my Gaussian example. I don't know how to get more specific than that. I offered to try a second example (the accelerometer). It would be more direct to your understanding of measurement, so would that work?

Then the model fails the test and is thrown out. No need for assumptions.

In my GR/QM example neither model has been thrown out.

Your measurement is still 3.2. The adjustment for noise is also not assumed, and is part of a model that is not assumed to be absolutely true.

I didn't repeat the conditions of the example, so maybe they got lost. As I said, this example is based on an actual test. In this instance, two different distributions fit almost equally well. There was no reason based in statistics for choosing one over the other (like my algebra analogy, where there was no reason to assume y rather than x). A Gaussian model was chosen because the math was easier that way. So, there was a reason for choosing a Gaussian model, but it had nothing to do with the test. Per my explanation of what I'm talking about - a choice with no binding reason - that constitutes an assumption.

Did you read any of the sources I referenced?

I have commented on your references. So where are the comments on mine? You just completely skipped over the very references that blow your argument out of the water.

I did comment. As I said, I skimmed them. I didn't see anything that changed what I was saying. My comments were not meant as a lecture, so I'm sorry to upset you. Rather it was a request. If there is something specific you want me to comment on, please quote it. For one of the articles, all I could get was the abstract.

I'm not trying to hide anything, so I will try one more citation on metrology to see if this is the type of thing you mean. This comes from The Metrology Handbook by J.L. Bucher. I am quoting from Chapter 27, Statistics, the section entitled "Fundamental Measurement Assumptions".

On page 277 of that section where he discusses autocorrelation, he begins with the statement, "An important test of the assumption that source data are normally distributed is to determine the autocorrelation coefficient for the data set."

That sounds like the type of thing you've been saying. However, a bit later he notes the result of the test is that "... the data set may be more confidently assumed to be normally distributed." My point is that the state has not been transformed from assumption to theory. It remains an assumption. All that these tests do (such as the paper you cited by Shlyakhter on nuclear constants) is increase confidence in the assumption.

In fact, upon reading Shlyakhter more closely, it appears to me that is what he is saying. After all, he concludes by saying, "These estimates seem to exclude all variants of nuclear constants change based on Dirac’s ‘Large Numbers Hypoth-esis'. It is, however, desirable to obtain as strict bounds on del E as possible."

 

[Resha Caner]

Also to your op, it's impossible for a theory to be completely unfalsifiable.

According to one of the great philosophers of science, Ernst Nagel, some theories can't be falsified.

 

[durangodawood]

He defined consciousness, which I didn't realize was an issue. It seemed to me the only reason he defined it was to make the point that the definition does not make it contingent on the material. Did he define "immaterial"? I don't recall seeing that. All he said was that it was a concept he could fathom. I've got no quarrel with Durango, but I don't know what concept of "immaterial" he is thinking about.

Yes I did all that^^^. Because I wanted to put some pressure on the positive-ontology litmus test idea.

 

[super animator]

According to one of the great philosophers of science, Ernst Nagel, some theories can't be falsified.

Than by definition is not science.

 

[durangodawood]

Than by definition is not science.

Which goes to show that there might be true - but scientifically unverifiable - statements.

 

[Loudmouth]

I offered the data for my Gaussian example.

That is a model, not an assumption.

In my GR/QM example neither model has been thrown out.

Both are understood to break down at different scales. They are only kept because they make good approximations at certain scales. Physicists have known for a long time that both are probably wrong in some way, and neither is assumed to be 100% true.

For that matter, we still use Newtonian physics to explain low speed and low gravity physics because it is a good approximation even though we know that Newtonian physics is wrong.

I didn't repeat the conditions of the example, so maybe they got lost. As I said, this example is based on an actual test. In this instance, two different distributions fit almost equally well. There was no reason based in statistics for choosing one over the other (like my algebra analogy, where there was no reason to assume y rather than x). A Gaussian model was chosen because the math was easier that way. So, there was a reason for choosing a Gaussian model, but it had nothing to do with the test. Per my explanation of what I'm talking about - a choice with no binding reason - that constitutes an assumption.

A choice with no binding reason is not an assumption. Never has been. You are thinking of the word "arbitrary", not assumption.

I did comment. As I said, I skimmed them. I didn't see anything that changed what I was saying. My comments were not meant as a lecture, so I'm sorry to upset you. Rather it was a request. If there is something specific you want me to comment on, please quote it. For one of the articles, all I could get was the abstract.

That is all you need. You claimed that it is assumed that physical laws do not change with time or change in location. I showed both of those to be false.

I'm not trying to hide anything, so I will try one more citation on metrology to see if this is the type of thing you mean. This comes from The Metrology Handbook by J.L. Bucher. I am quoting from Chapter 27, Statistics, the section entitled "Fundamental Measurement Assumptions".

Which of those assumptions can not be tested?

On page 277 of that section where he discusses autocorrelation, he begins with the statement, "An important test of the assumption that source data are normally distributed is to determine the autocorrelation coefficient for the data set."

So not an assumption, but a testable proposition.

That sounds like the type of thing you've been saying. However, a bit later he notes the result of the test is that "... the data set may be more confidently assumed to be normally distributed." My point is that the state has not been transformed from assumption to theory. It remains an assumption. All that these tests do (such as the paper you cited by Shlyakhter on nuclear constants) is increase confidence in the assumption.

It remains a tentative conclusion, no matter how awkwardly the author writes it. No statitistical model is assumed to be 100% true. It is similar to "assuming" that Newtonian physics is true for low speed and low gravity physics, even though we KNOW that it is not true.

In fact, upon reading Shlyakhter more closely, it appears to me that is what he is saying. After all, he concludes by saying, "These estimates seem to exclude all variants of nuclear constants change based on Dirac’s ‘Large Numbers Hypoth-esis'. It is, however, desirable to obtain as strict bounds on del E as possible."

So further testing is needed on the things you claim no one has tested.

 

[Resha Caner]

Than by definition is not science.

Which goes to show that there might be true - but scientifically unverifiable - statements.

Not that I disagree with either statement - at some point this becomes a philosophical discussion, though that boundary can be hard to define. But, if there is an intended subtext of "and therefore the point is unimportant", then I would disagree.

Specifically, Nagel's classic example regards Newton's second law (F = ma). He claimed it can't be falsified ... and note he was quite aware of the Einstein etc. corpus of work when he made that claim ... further note his claim was not an attempt to retain Newtonian mechanics.

 

[Davian]

Not that I disagree with either statement - at some point this becomes a philosophical discussion, though that boundary can be hard to define. But, if there is an intended subtext of "and therefore the point is unimportant", then I would disagree.

Specifically, Nagel's classic example regards Newton's second law (F = ma). He claimed it can't be falsified ... and note he was quite aware of the Einstein etc. corpus of work when he made that claim ... further note his claim was not an attempt to retain Newtonian mechanics.

The phrase that comes to my mind regarding unfalsifiable statements is that they are "without significance".

 

[Resha Caner]

Physicists have known for a long time that both are probably wrong in some way, and neither is assumed to be 100% true.

Why are these models not 100% true? Is the math wrong? Were the measurements bad?

A choice with no binding reason is not an assumption. Never has been. You are thinking of the word "arbitrary", not assumption.

Are you saying the meaning of assumption differs between math and science? Or are you saying that there is a "correct" choice for the parallel postulate? Basically, what I'm asking is: What is your definition of an assumption?

You claimed that it is assumed that physical laws do not change with time or change in location. I showed both of those to be false.

No, I said that after all known compensations are applied (temperature, pressure, etc.), it is assumed the remaining changes (time, location, etc.) do not affect the measure. My example, then, was: I assume that I can measure a wall in one place, cut the board in a different place, and the measure is applicable in both places.

So not an assumption, but a testable proposition.

You've chosen to call the author awkward and change his choice of words. I thought he was quite clear.

So further testing is needed on the things you claim no one has tested.

Shrug. If that's possible. It seems that's how Shlyakhter left it as well.

Which of those assumptions can not be tested?

When an assumption of normal distribution and cosine distribution correlate equally well, a test will not determine which to use. In fact, the cosine distribution has been suggested as an alternative. See "Some Useful Alternatives to the Normal Distribution" by Victor Chew, The American Statistician, June 1968.

Should I conclude that my suggestion for meeting your request for more examples by discussing accelerometers has not been accepted? You want to stay with the Gaussian example?

 

[Loudmouth]

Why are these models not 100% true?

BECAUSE SCIENCE IS TENTATIVE!!!

How many times do we need to tread over the same territory?

Are you saying the meaning of assumption differs between math and science?

YES!!!! ABSOLUTELY!!!! You finally have something right. In math, equations are not tentative and are metaphysically true. Assumptions are just that, dogmatic statements that are assumed to be true.

This doesn't happen in science.

What is your definition of an assumption?

An assumption is a statement that is accepted as 100% true without evidence or testing.

No, I said that after all known compensations are applied (temperature, pressure, etc.), it is assumed the remaining changes (time, location, etc.) do not affect the measure.

And I showed that this is false. Changes in the basic laws of physics are tested to see if they remain constant over time and location. It is not assumed.

You've chosen to call the author awkward and change his choice of words. I thought he was quite clear.

It was clear that they are using the word assumption where it doesn't apply.

When an assumption of normal distribution and cosine distribution correlate equally well, a test will not determine which to use. In fact, the cosine distribution has been suggested as an alternative. See "Some Useful Alternatives to the Normal Distribution" by Victor Chew, The American Statistician, June 1968.

So you make an arbitrary decision of which mathematical model to use. That is being arbitrary.

Also, these are statistical models, not assumptions. People can accept or reject your conclusions based on whether they think the model applies or not. The measurements themselves exist outside of the Gaussian model as well.

Should I conclude that my suggestion for meeting your request for more examples by discussing accelerometers has not been accepted? You want to stay with the Gaussian example?

[/quote]

Accelerometers are fine.

 

[Upisoft]

That's more of a hypothesis than a theory.

Well, multiverse theory is called theory as well. Sometimes naming conventions play jokes on us.

 

[Resha Caner]

BECAUSE SCIENCE IS TENTATIVE!!!

As just a quick check on the definition of "tentative", I looked here: Tentative | What is the Definition of Tentative? | Dictionary.com

It uses the word "uncertain" in that definition. As such, your statement comes across to me as "Science is uncertain because it is uncertain."

Hence I am asking why. Why is it tentative?

YES!!!! ABSOLUTELY!!!! You finally have something right. In math, equations are not tentative and are metaphysically true. Assumptions are just that, dogmatic statements that are assumed to be true.

This doesn't happen in science.

I wasn't saying I think math & science treat assumptions differently. I was asking if you thought that. To be honest, I can't make much of this. I asked about mathematical assumptions, not equations. Further, I'm not sure I understand your usage of "dogmatic", because I find mathematicians to be quite the opposite.

Regardless, it seems you think mathematical and scientific assumptions to be different. So, let me ask, is science dependent upon mathematics?

An assumption is a statement that is accepted as 100% true without evidence or testing.

OK. So it is possible to test all scientific assumptions and convert them to theories? There is a 100% certainty that they can all be tested?

And I showed that this is false. Changes in the basic laws of physics are tested to see if they remain constant over time and location. It is not assumed.

I don't think that's what your cited papers said. For example, the paper on nuclear constants said it was studying the effects of expansion of the universe. I don't see that as saying time or location is the cause. For similar expansion, I would think the effect is the same regardless of when or where it happens. So this fits with what I said. Once the effect is known (be it expansion, gravity, or dark energy), the compensation is added to the measure. Yet there still remains an assumption that those compensations which have not been added are not affecting the measurement.

But suppose, for the sake of the discussion, a time effect were found. And? It gets added to the list of compensations, and we still have the assumption that effects not in the compensation list are not affecting the measurement.

So you make an arbitrary decision of which mathematical model to use. That is being arbitrary.

And statistics never effect your measurement? It has no impact on sample size, and sample size has no impact on the measurement? Does other signal processing ever effect the measurement? Filters, A to D converters, anything?

I guess you need to tell me exactly what you consider the "measurement" to be.

Accelerometers are fine.

Just to insure we're on the same page: Accelerometer - Wikipedia, the free encyclopedia

So, following up on my previous comment. What is the "measurement" in an accelerometer?

 

[Mr Clean]

OK, I'll try. There are exactly 5000 universes, but there is no way to travel between them.
Falsify this.

There are exactly 5000 universes, but there is no way to travel between them...unless you eat tacos.

?