Is "Equal a priori Probabilities" the correct scientific approach?

[J-H-C]

Hello,
My understanding of the scientific approach when faced with an unknown sequence:
Hypothesis: There is an order/signal/bias/code = Goal to achieve
Null hypothesis: Absence of bias/order/code = Equal Probability = Target to destroy.

Science tries to break the code.
In order to prove the Null wrong, one designs a study and analyzes a random sample of the sequence to look for any pattern, order, code, signal, meaning or predictability etc. If meaning or order is found then the proof is shown and Null is destroyed if no order is found then Null stands till destroyed by someone else.

But by building theories like Entropy and Evolution etc on Equal a priori Probabilities scientific community has created a serious conflict of interest with finding the code because if someone was to break the code that would mean re-writing all science books and it would prove scientist wrong and naive to have built laws of Physics assuming equal probabilities.

As I see it Equal a prior Probabilities is as anti-scientific of an approach as it can theoretically get as:
1. It creates a bias against finding bias/order
2. It creates a blind spot in central vision as it can be used to explain "all order"
3. It is not falsifiable
4. conclusion made on it are very misleading and wrong if it turns out Equal a priori Probabilities did not exist
5. It should never be used as foundation to build hypothesis unless systematic studies failed to show any trace of discernible order or bias
6. It is assumed as a position of failure when one fails to find any meaning/signal/bias in the data
7. "Assumed absence of bias/order" is as weak of a platform to build theories upon as it can get
8. It hinders scientific discovery of order and bias

Now the most important question:
Does equal probability even exist? It has not be proven true. If it turns out to be false then obviously science books would need to be re-written.

And as I see it, Gravity imposes "an order" on all matter and creates "a bias" which makes matter move favouring one direction over all others and with that bias Gravity arranges matter in "an order" by "Bubble sort" method.
Example 1. all stars have the same order with Hydrogen being the outer most layer and then Helium etc.
Example 2. Ships sail because of matter being Bubble sorted by Gravity
Example 3. Rivers flow as water molecules move according to the bias created by Gravity
Example 4. Hot air rises
Example 5. There is more pressure at the bottom of the air column than at the top.

As there is a bias present in the universe hence it destroys the assumption of equal probability/absence of bias.
But obviously acknowledging this "bias" would mean re-writing science books and hence this bias along with all the order it creates in the universe has been ignored as insignificant and blamed it on rare chance. So my above analysis that Equal a priori Probabilities creates a bias and blind spot against bias is proven true.
Truth is science cannot progress unless it does away with Equal a priori Probabilities as it is creating an illusion of knowledge and is preventing scientists from seeing or even looking for the order/bias IMO.
Who is supposed to be looking for signal if scientists are trying to prove it's all noise?

What is the right scientific approach in your opinion?
1. To look for bias/order
or
2 . Try to prove it is all noise

Peace

 

[Dale]

I think that this post shows a rather large misunderstanding about the role of null hypothesis significance testing (NHST) and the null hypothesis (NH), what you are calling “equal a priori probabilities”.

First:

by building theories like Entropy and Evolution etc on Equal a priori Probabilities

Modern theories are not built on the NH. Quite the opposite, they are built on rejecting it through NHST.

It creates a bias against finding bias/order

No, it does not create bias, but it does require a minimal burden of evidence to be presented before a scientific proposition will be accepted. This is correct scientific reasoning. You do not want to accept claims without evidence to support them, that would be unscientific.

It is not falsifiable

The NH is in fact falsified in the vast majority of published papers.

conclusion made on it are very misleading and wrong if it turns out Equal a priori Probabilities did not exist

Yes. Both type I and type II errors are misleading and wrong.

It should never be used as foundation to build hypothesis unless systematic studies failed to show any trace of discernible order or bias

This is in fact a valid criticism of NHST. You cannot use standard statistical techniques to provide evidence in support of the NH. Luckily there are alternative approaches, such as Bayesian statistics which can be used to rigorously provide persuasive evidence in support of the NH when scientifically necessary.

It is assumed as a position of failure when one fails to find any meaning/signal/bias in the data

See the previous point.

"Assumed absence of bias/order" is as weak of a platform to build theories upon as it can get

There are indeed substantial limitations to NHST, but I don’t think that you have actually touched on any of them. The scientific community is increasingly aware of these problems which are actually quite subtle and difficult to resolve. I don’t believe that your criticisms are relevant to the actual issues that do exist.

It hinders scientific discovery of order and bias

Actually, the exact opposite is true. It is too easily fooled into accepting a false discovery. This is known as the replication crisis.

Truth is science cannot progress unless it does away with Equal a priori Probabilities

I think there is a growing scientific movement to replace the NHST methodology, or at least to supplement it with other statistical techniques that will serve to address the replicability crisis and other related scientific weaknesses of NHST.

 

[J-H-C]

Thanks for replying. But your reply has added to my confusion. Please correct me if I am wrong as I am not a statistician nor a physicist.
Statistically speaking what exactly is Evolution hypothesis trying to prove?
Is it trying to prove that the orders we are observing (DNA codes) are a result of random process (absence of bias) or
is it trying to "prove presence of order and possible bias" to explain the order?
As I see it, it is trying to prove absence of order/bias (Null) and not the alternative hypothesis as it is built on the assumption of absence of bias (equal probability).

In other words has Evolution mislabeled the Null and Alternative hypothesis ?

And same question applies to Entropy and the current scientific understanding of the universe too, why assume absence of bias (Equal a priori Probabilities) to build conclusions!
Am I missing something?
I appreciate your time.
Peace

 

[Dale]

Statistically speaking what exactly is Evolution hypothesis trying to prove?

Statistically, the “evolution hypothesis” is that genes possessed by individuals who fail to reproduce become less common in a population and genes possessed by individuals who successfully reproduce become more common.

A typical experiment might be to collect a population of fruit flies containing a gene for resistance to a toxin and another for susceptibility to the toxin. Then dividing the population into two samples, one exposed to the toxin and one not exposed to the toxin. The null hypothesis would be that the frequency of the resistant gene would not increase in the toxin-exposed population compared to the toxin-naive population. A substantial body of research had falsified that null hypothesis.

As I see it, it is trying to prove absence of order/bias (Null) and not the alternative hypothesis as it is built on the assumption of absence of bias (equal probability).

I don’t know what to say other than your assessment is incorrect. See above. You seem to be bringing in some unrelated concept, but I am not sure what it is.

In other words has Evolution mislabeled the Null and Alternative hypothesis ?

No. It is odd to even think that it is possible for a theory to mislabel a hypothesis. A theory does not have a null hypothesis per se. A hypothesis is part of an experiment. So you set up an experiment to test a theory and as part of that experiment you define a hypothesis. The null hypothesis is the hypothesis that there is no effect from the experimental treatment. The hypothesis (null and otherwise) is less a part of the theory and more a part of the experiment.

And same question applies to Entropy

Entropy is not a theory. It is a quantity. I think you are referring to the second law of thermodynamics. I don’t see how you could possibly assume that the second law of thermodynamics is assuming the null hypothesis.

 

[hutchphd]

Am I missing something?

I think you are but my apologies in advance if I misunderstand.

The basis for most experimental analysis assumes that absent a causal link, the graph of one quantity vs another should look like "snow" (old analog TV reference). That is the absence of bias...and it seems to make sense.

You could assume instead that you expect such an "acausal" graph to always show an image of Mickey Mouse. Not seeing his image would imply some connection.

It appears that this Disney world is not where we live.

So I ask you what is it you propose to replace the assumption of equal a priori probabilities??

 

[Dale]

@J-H-C I note that your objection to NHST seems to be based, not on any actual flaw in NHST (of which there are many), but rather on some vague connection with politically charged topics. I would recommend looking at the actual methodological issues, and not the connection to political issues.

Your only valid methodological criticism is number 5, so perhaps we should start there? Or perhaps you might prefer to discuss why one of your other numbered criticisms is not valid?

 

[J-H-C]

Thank you.
I like the example of "snow" and that is what I was calling "noise"
Let the explain my point again this way. A DNA code of an ant is obviously anything but "snow". It is a long book of science which can and is translated into another product. So an ant is one book written in two languages. It is as good of a proof against snow as you can get IMO.
If the goal is to find evidence of order then the probability (p value) that an ant DNA sequence could be result of a random process is very low and hence it should prove absence of snow so to speak.
In other words, it is much much more likely that an ant DNA code is result of some bias (unknown) then equal probability as I see it.
To me it appears that something is knowingly writing these codes and puts in place mechanism for their translation too. If you look up the definition of "random" then absence of meaning is a requirement for something to be a random process.
But because "it's all snow" assumption, even overwhelming evidence of order, meaning and intent is being attributed to rare chance in order to defend "its-all-snow theory".
As you know all order can be attributed to a probability if equal probabilities is assumed so how to do you falsify Evolution unless you prove equal probabilities do not exist.
So question is, do equal probabilities exist?
I see Gravity creating a relentless bias for all molecules which makes them face and move preferring one direction over all others and hence it destroys equal probability for everything. Because of Gravity the die has one side which is lot bigger than others so how can there be equal probabilities?
I am proposing that we realize that equal probabilities do no exist and then try to understand how the bias of Gravity explains all creation of order and phenomenon we are seeing.
How can there be "snow" in presence of Gravity?

 

[J-H-C]

Thanks Dave,
What I was trying to say is that Science has an agenda which is "to know" if there is a meaning/code /order or bias in something which was unknown or unpredictable before.

But as I see it somehow the goal of scientists has become to "prove absence" of order or bias and even that without negative systematic studies. Why?
Isn't "proving absence" an oxymoron? What evidence do you have to look for to prove "assumed absence" of something? Answer: Nothing!
I think you would agree that it will be naive to ignore the possibility of personal biases and prejudices of scientists at play.
We should make sure that goals of science and scientists are aligned and there is no conflict of interest as we are all humans and somewhat blind to our own prejudices and we want to prove the reality the way we like it to be.
For example:
Q. Is an ant DNA code created by an author (bias) or not?
What science would like to know may be different than what a certain scientist would want to prove.
Hence I am questioning the current scientific approach as it seems to be geared towards proving absence of order/bias rather than looking to "prove presence" of order/bias as it should be.
Peace

 

[Dale]

In other words, it is much much more likely that an ant DNA code is result of some bias (unknown) then equal probability as I see it.

Obviously, but I am not sure why you think that statement represents an objection to NHST.

Frankly, it appears that you have a personal dislike for some typical political “hot button” topics and are transferring that dislike of the topics into a rather unfounded objection to the NHST methodology. I would again recommend focusing on the methodological issues.

But because "it's all snow" assumption, even overwhelming evidence of order, meaning and intent is being attributed to rare chance in order to defend "its-all-snow theory".

This does not even bear a passing resemblance to the theory of evolution.

 

[Dale]

Thanks Dave,

There is no Dave in this conversation, so I assume you meant me. It would be easier to tell if you use the quote feature.

Science has an agenda

Science is a tool. It has no more agenda than a screwdriver or a hammer. People have agendas, including scientists.

But as I see it somehow the goal of scientists has become to "prove absence" of order or bias and even that without negative systematic studies. Why?

Can you post a specific reference to a specific experiment where a specific scientist attempted to use NHST to “prove absence of order or bias”? I am sure it has happened, but it is quite rare, so it might be good to have a concrete example to discuss. I certainly do not think this is a broad ranging issue, and you may have a hard time finding such an example.

What evidence do you have to look for to prove "assumed absence" of something? Answer: Nothing!

Which is precisely the reason why science is not generally done this way. See: https://www.physicsforums.com/threa...othesis-for-biased-coins.950744/#post-6140788

Hence I am questioning the current scientific approach as it seems to be geared towards proving absence of order/bias rather than looking to "prove presence" of order/bias as it should be.

I am questioning your characterization of “the current scientific approach”. I think you will be very hard pressed to actually find more than a minuscule handful of examples where a standard NHST was actually used as evidence to support the null hypothesis. So what you describe is far removed from the actual “current scientific approach”

 

[Dale]

As this thread has turned away from an actual discussion about the methodological issues of NHST and into an uninformed referendum on unpopular scientific “hot topics” it is time to close it.

@J-H-C if you find a reference of a specific example of using NHST to support the null hypothesis or a paper discussing some of the actual issues with NHST or the replication crisis then please start a new thread with that reference in the OP, and refrain from making unsubstantiated claims about general controversial topics.