[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/sci/ - Science & Math


View post   

File: 50 KB, 742x609, AbsenceOfEvidence.gif [View same] [iqdb] [saucenao] [google]
9535327 No.9535327 [Reply] [Original]

What does /sci/ think about absence of evidence?

>> No.9535329

I don't know enough about it

>> No.9535331

This troll has gone long enough, but autists here will still respond.

>> No.9535334

>>9535331
What troll?

>> No.9535360

>>9535327
Definition 2 is wrong. Absence of evidence is not "not A" it's "A = null"

>> No.9535368

>>9535327
I think it's not evidence of absence but it's evidence of an alternative.

>> No.9535376

>>9535360
>Definition 2 is wrong. Absence of evidence is not "not A" it's "A = null"
This. The "evidence" in Bayes theorem is the event in the conditional, and so "not A" being placed there means there is in fact evidence being considered.

>> No.9535491

- If A -> B, then ~A -> ~B is fallacious. You could can say ~B -> ~A
quad porno democratic

>> No.9536432

>>9535327
absence of evidence is always compelling, when some attempt has been made to gather it

>we predict a particle should exist with this energy signature
>look and don't find it
>ABSENCE OF EVIDENCE ISN'T EVIDENCE OF ABSENCE MY THEORY IS TRUE WE CAN'T KNOW NOTHING
basically

>> No.9536450

>>9535360
This. Fuck off with this bullshit OP

>> No.9536586

>>9535327
Depends on your definition of evidence

https://en.wikipedia.org/wiki/Raven_paradox

>> No.9536588
File: 107 KB, 645x729, brainlet35.png [View same] [iqdb] [saucenao] [google]
9536588

>Definition: If there is evidence for X, then the probability of X is greater than if there weren't.
>Theorem: If there is no evidence for X, then the probability of X is smaller than if there were.
Bayesianists are retarded.

>> No.9536590

>>9535327
>P(B|A) > P(B|~A) <=> 1 - P(~B|A) > 1 - P(~B|~A)
lol

>> No.9536605

>>9535360
>A = null
No, that would mean that no possible event could be evidence. Absence of evidence on the other hand is that the possible event that would be evidence did not occur.

For example, an intelligent radio signal from outer space would be evidence of alien life, but this signal is currently absent. Saying that evidence of alien life = null is to say that the probability space does not contain the event of receiving an alien radio signal.

>> No.9536608

>>9535376
>The "evidence" in Bayes theorem is the event in the conditional
This ignores that the phrase refers to specific evidence for an event. A conditional is not evidence for all events.

>and so "not A" being placed there means there is in fact evidence being considered.
Which is correct since the absence of evidence for one thing is evidence for something else.

>> No.9536615

>>9536432
Mmm well I would say it's a bit stronger than that. You don't necessarily need to look for evidence to have it. It could just fall into your lap. As long as that possibility exists, its absence is evidence of absence.

>> No.9536661

>>9535327
Conditional probabilities and Bayes' Theorem don't work if the conditioning probability is 1 or 0 since in this case you are using both A and A*.

Evidence either exists somewhere or it doesn't. P(A) = 1 or 0, nothing inbetween. As others have said, this means A* is the null set.

You can't just fix this by saying 0 < P(A) < 1, because what does that mean?

>> No.9536677

>>9536661
>using a frequentist interpretation of probability with bayes' theorem
lelllllllllllllllllllll

>> No.9536752

>>9536661
>Evidence either exists somewhere or it doesn't. P(A) = 1 or 0
Wew lad, nice bait.

>> No.9536757

>>9536752
Right, since "either it exists or it doesn't" means P(A) = 0.5 obviously.

>> No.9536766

>>9536757
No, it doesn't necessarily mean that either. It might mean that in certain cases. For example if you flip a fair coin then the chance it landed heads is 0.5 before you look at the result. Not "1 or 0." But again this is clearly bait as anyone who has passing knowledge of what a probability is would know this.

>> No.9536813

>>9536677
>>9536752
>>9536766
Not baiting. If you think I'm wrong about what the probability of A is, come up with a definition of A is - what does it represent - that doesn't contradict the manipulations that supposedly follow.

Suppose P(A) = 0.4. What does this mean? 40% chance of evidence existing? That's a meaningless statement. 40% of finding evidence given that it exists? That's just kicking the can down the road; how would you evaluate such a probability without presupposing the conclusion the argument?

>> No.9536817

You assumed your conclusion in Definition 1. People who say that phrase do not use it when Definition 1 applies.

>> No.9536830

>>9535327
nice. too bad evidence is more than one probability being greater than the other.

there's no evidence of your dick. therefore you have no dick.

>> No.9536844

>>9536813
>If you think I'm wrong about what the probability of A is, come up with a definition of A is - what does it represent - that doesn't contradict the manipulations that supposedly follow.
The entire definition is right there. A is an event which increases the probability of B when true.

>Suppose P(A) = 0.4. What does this mean? 40% chance of evidence existing?
Does "evidence existing" mean that the chance of B increases? If yes, then yes it means that.

>40% of finding evidence given that it exists?
I don't see what this has to do with the problem. Where do you see the conditional "given that it exists"?

>> No.9536847

>>9535327
As far as I can make out, it's just mincing words. Logical negation is not a good representation of what "absence" means.

>> No.9536850

>>9536813
>what is the likelihood we'd see ten heads in a row, given the coin is fair
>what is the likelihood the coin is fair, given we saw ten heads in a row
what oh what could it all mean, I guess everything is either P(X) = 1 or P(X) = 0 because we talk personally to god who has the ultimate set of tools, how does he even deal with the absence of tails it's truly a puzzler

>> No.9536857

>>9536847
>we have no evidence of aliens existing
Have you looked?

A) Yes. Then absence of evidence is evidence of absence.
B) No. Then look you fucking idiot.

>people actually take a class to learn this

>> No.9536863

>>9536857
how do you know if you've fully looked

>> No.9536866

>>9536857
>>we have no evidence of aliens existing
Who are you quoting?

>> No.9536868

>>9536857
>Have you looked?
What exactly does this mean in mathematical terms?

>> No.9536876

>>9536830
>too bad evidence is more than one probability being greater than the other.
Such as?

>> No.9536879

>>9536868
It means P(A) > 0

>> No.9536881

>>9536866
I'm presenting a scenario and using the greentext to highlight a different voice in a hypothetical conversation.

>> No.9536882

>>9536879
>It means P(A) > 0
Then why does "A) Yes. Then absence of evidence is evidence of absence" follow?

>> No.9536883

>>9536882
The pic in the OP explains that.

>> No.9536885

>>9536863
You don't, which is why it's a question of probability.

>> No.9536895

>>9535327
Evidence should be defined as P(B|A) > P(~B|A). The way you write it, P(B|A) could be 0.00001 and A could still be considered evidence for B.

>> No.9536901

>>9536895
So you can't have evidence that an unlikely event happened?

>> No.9536929

>>9536847
>As far as I can make out, it's just mincing words. Logical negation is not a good representation of what "absence" means.
This. The pic in OP seems to be mixing up "absence of evidence" and "having complementary evidence".

>> No.9536932

>>9536844
You're completely and utterly missing the point. Of course that's how probability works, in terms of events. The problem is there's no way to map an event occurring to the question of existence/absence in this way. That's why I asked what A refers to, not about its implicit mathematical definition in the inequality that follows.

>> No.9536940 [DELETED] 

>>9536901
Think about how conditional probabilities work.

>> No.9536941

>>9535327
If P(b|a) > P(b| not a) then doesn't that mean a is by definition evidence, and so can not be an absence of evidence?

>> No.9536947

>>9536941
also on this point, wouldn't it make more sense for the "absence of evidence" to be the union of all events (not A) such that P(B|A)>P(B|not A)?

>> No.9536997

>>9536932
It's pretty simple. Either evidence is present and increases the chance of B, or it is absent. That's exactly how it is defined in the OP. Either present a counterargument or accept it.

>> No.9537009

>>9536997
>Either evidence is present and increases the chance of B, or it is absent.
What does that mean mathematically? The evidence being used are just events (sets), what does it mean for a set to be "present" or "absent"?

>> No.9537015

>>9536941
By that logic there is no such thing as an "absence of evidence," since absence of evidence for B is evidence against B.

>> No.9537019

>>9537015
>By that logic there is no such thing as an "absence of evidence," since absence of evidence for B is evidence against B.
Why does absence of evidence for B being evidence against B mean that absence of evidence doesn't exist?

>> No.9537038

>>9537009
>What does that mean mathematically?
Again, it's written mathematically in the OP. Why are you asking questions that have already been answered?

>The evidence being used are just events (sets), what does it mean for a set to be "present" or "absent"?
Again it's clearly defined in the OP. Evidence is present when the event defined as the condition which increases the chance of B occurs. It is absent when the event does not occur.

>> No.9537041

>>9535327
Definitions 2 and 3 are conflicting.

>> No.9537046

>>9537038
>It is absent when the event does not occur.
So how do you distinguish when A "does not occur" and when not A occurs?

>> No.9537057

>>9537019
Why are you asking me to explain your own argument?

>If P(b|a) > P(b| not a) then doesn't that mean a is by definition evidence, and so can not be an absence of evidence?

Any event a which is an absence of evidence for B is evidence for b. Thus there cannot be an absence of both evidence for B and evidence for b.

>> No.9537062

>>9537041
Wrong.

>> No.9537067

>>9535327
A better definition for absence of evidence for B would be an event A such that P(B|A) = P(B).

>> No.9537071

>>9537046
Why do you think they need to be distinguished?

>> No.9537074

>>9535327
The Absence of Evidence is just Unfalsifiability

>> No.9537077

>>9537071
>Why do you think they need to be distinguished?
Because they are two separate cases.

>> No.9537080

>>9537057
>Why are you asking me to explain your own argument?
I'm not, you're the one who made the claim.

>> No.9537105

>>9537067
That would mean that the presence of evidence against B is the presence of evidence for B.

>> No.9537108

>>9537077
Why?

>> No.9537112

>>9537105
I don't follow, can you elaborate?

>> No.9537118

>>9537080
Did you read the post before responding?

>> No.9537129

>>9537112
Let event C be the presence of evidence against B such that it decreases the chance of B.

If the absence of evidence for B would be an event A such that P(B|A) = P(B), then event C occurring means that there is not an absence of evidence for B, since it violates the definition.

Thus C means that evidence against B is present and it means that evidence for B is not absent, i.e. evidence for B is present.

>> No.9537132

>>9537118
>Did you read the post before responding?
Yes, you wrote "By that logic there is no such thing as an "absence of evidence," since absence of evidence for B is evidence against B." and I asked you "Why does absence of evidence for B being evidence against B mean that absence of evidence doesn't exist?". What part of that was my own argument?

>> No.9537134

>>9537132
I don't see any reference to the last post you were responding to. Did you read it?

>> No.9537138

>>9537134
>I don't see any reference to the last post you were responding to. Did you read it?
Yes, I quoted it in full.

>> No.9537141

>>9537108
>Why?
For example, the probability of the throw of a die giving 6 is different than the probability of the throw of a die giving 6 given you rolled an odd number.

>> No.9537145

>>9535327
but probability of B given A is not higher then probability of B given negative A.B is independent of A.
i dont think whoever wrote that took\understood his probability course .

>> No.9537146

>>9537129
>If the absence of evidence for B would be an event A such that P(B|A) = P(B), then event C occurring means that there is not an absence of evidence for B, since it violates the definition.
Can you rephrase this part? It's not clear to me what you're trying to say. What exactly is violating which definition?

>> No.9537185

>>9537138
No you didn't. Read >>9537057 in full before responding.

>> No.9537194

>>9537141
I don't see what that has to do with distinguishing A not occurring and not A occuring. Are you saying they are mutually exclusive?

>> No.9537200

>>9537194
>I don't see what that has to do with distinguishing A not occurring and not A occuring.
" the probability of the throw of a die giving 6" and "the probability of the throw of a die giving 6 given you rolled an odd number" are two different probabilities, and so it makes sense to distinguish them.

>Are you saying they are mutually exclusive?
No.

>> No.9537201

>>9537185
>No you didn't.
Which part did I not include in the quote?

>> No.9537202

>>9537145
>but probability of B given A is not higher then probability of B given negative A.B is independent of A.
By definition B is dependent on A. I don't think you have ever passed a probability course.

>> No.9537205

>>9537146
P(B|C) =/= P(B) therefore C is not an absence of evidence of B.

>> No.9537208

>>9537201
Nothing you quoted was even in that post. Just stop posting.

>> No.9537211

>>9537205
>P(B|C) =/= P(B) therefore C is not an absence of evidence of B.
Why is that an issue? Not everything has to be an absence of evidence of B.

>> No.9537212

>>9537208
>Nothing you quoted was even in that post.
Which part did I not include in the quote?

>> No.9537216

>>9535327
https://www.youtube.com/watch?v=_w5JqQLqqTc

>> No.9537237

>>9537211
The non-absence evidence of B is the presence of evidence of B.

>> No.9537242

>>9537237
>The non-absence evidence
How is this defined?

>> No.9537260

>>9535327
I dont get it. So | is NAND? P(B|A) > P(B|notA) is true for values only true for second option: A=0, B=1, right: 10 01 00 11. Then P(B|A) > P(B|notA) mean 1|0 > 1|1 and is true for those values, right? Then everything that goes along is inconsistent because 1-1 is not bigger then 1-1. Explain these things for me, i see this P() thing for the first time

>> No.9537261

>>9537242
The non-absence of evidence for B is the presence of evidence for B.

>> No.9537268

>>9537261
>The non-absence of evidence for B is the presence of evidence for B.
So what does this have to do with not everything having to be an absence of evidence of B? I just don't see why some definition is apparently being violated

>> No.9537365

>>9537260
https://en.wikipedia.org/wiki/Conditional_probability

>> No.9537370

>>9537268
If P(B|C) =/= P(B) then C occurring means evidence for B is present. But C is the presence of evidence against B. Thus we have a contradiction that invalidates your definition

>> No.9537377

>>9537370
>If P(B|C) =/= P(B) then C occurring means evidence for B is present.
Yes.

>But C is the presence of evidence against B.
Why does this follow?

>> No.9537382

>>9537377
>Yes.
By this I mean either C or not C is evidence for B.

>> No.9537384

>>9537377
C was defined as such from the beginning...

>> No.9537391

>>9535327

Absence of evidence is evidence of absence, it just isn't proof of it.

>> No.9537394

>>9537384
>C was defined as such from the beginning...
If C is defined to be evidence against B then why does it matter if C is not absence of evidence for B?

>> No.9537396

>>9537382
The non-absence of evidence for B is its presence. Whatever you define as the absence of evidence for B, its complement is presence of evidence for B. Since your definition does not include evidence against B, it falls.

>> No.9537403

>>9537394
Because if it's not absence it's presence. Absence and presence of something are mutually exclusive and collectively exhaustive.

>> No.9537404

>>9537396
>Whatever you define as the absence of evidence for B, its complement is presence of evidence for B.
Can you elaborate?

>Since your definition does not include evidence against B, it falls.
What do you mean by "it falls"?

>> No.9537406 [DELETED] 

>>9537404
See >>9537396

>> No.9537411

>>9537404
See >>9537403

>> No.9537413

>>9537406
>See >>9537396
That doesn't explain why "Whatever you define as the absence of evidence for B, its complement is presence of evidence for B". Obviously your statement is only true if whatever you've defined as the absence of evidence for B is the complement of presence of evidence for B. Why is it so?

>> No.9537417

>>9537403
>Because if it's not absence it's presence.
If what is not absence then it's presence?

>> No.9537418

>>9537413
You mean why can't something be both present and absent? It follows from the definitions of absence and presence.

>> No.9537420

>>9537417
C.

>> No.9537421

>>9537418
>You mean why can't something be both present and absent?
No, I asked why the definitions must be the way you said they must be.

>> No.9537423

>>9537420
>C.
What does it mean for C to be presence? I thought C was evidence.

>> No.9537424

>>9537421
Because these words have a certain meaning in the English language.

>> No.9537427

>>9537423
Refer back to >>9537129 where C is defined as the presence of evidence against B.

>> No.9537428

>>9537424
>Because these words have a certain meaning in the English language.
English is ambiguous which is why we're using mathematical terms. For the same reason that "evidence" and "absence" have certain meanings in English yet are still being redefined here in terms of mathematics.

>> No.9537434

>>9537427
>Refer back to >>9537129 where C is defined as the presence of evidence against B.
You've already said that C "is presence", telling me you've defined it as such doesn't explain what that means. What does it mean for C to be presence?

>> No.9537441

>>9537428
>English is ambiguous which is why we're using mathematical terms.
I don't see any ambiguity in this context. You have not presented an argument why presence and absence of something should not be considered mutually exclusive and collectively exhaustive, when that is how they are used by English speakers. You can define absence mathematically however you want, but if your model diverges from what the word absence implies then it isn't a useful model, certainly not for interpreting a phrase containing the word absence. If you want to talk about some different concept from which the phrase refers, then I don't see the point in paying attention to your posts.

>> No.9537444

>>9537441
>I don't see any ambiguity in this context.
The ambiguity is that you're committing the very fallacy you're denying. You've assuming that the phrase "absence of evidence" ought to be modeled mathematically as a construction that actually models presence of evidence to the contrary.

>> No.9537445

>>9537434
>You've already said that C "is presence", telling me you've defined it as such doesn't explain what that means.
C is the presence of evidence against B. If you refer back to >>9537129 I said exactly what it means, it decreases the chance of B.

>> No.9537454

>>9537445
>C is the presence of evidence against B.
This is just more ambiguity, since now we have that "evidence against B" decreases the probability of B and so does "presence of evidence against B". You're just mincing words.

>> No.9537457

>>9537444
It seems to have been modeled quite successfully since I've been able to easily shut down every counterargument you attempt against it. You can cry about it all you want, but until you actually provide an argument you're just embarrassing yourself.

>> No.9537458

>>9537457
>It seems to have been modeled quite successfully since I've been able to easily shut down every counterargument you attempt against it. You can cry about it all you want, but until you actually provide an argument you're just embarrassing yourself.
You haven't actually said anything substantial here. Could you represent your argument symbolically? It seems very muddled.

>> No.9537464

>>9537454
>This is just more ambiguity, since now we have that "evidence against B" decreases the probability of B and so does "presence of evidence against B"
Obviously evidence against B doesn't decrease the probability of B if it's not present. If you can't handle a little shorthand then I suggest you go play somewhere else.

>You're just mincing words.
Stop projecting.

>> No.9537466

>>9537464
>Obviously evidence against B doesn't decrease the probability of B if it's not present.
Of course it does, otherwise it wouldn't be evidence against B. Rolling an odd number on a die decreases the probability of a 6 irregardless of whether it's "present".

>> No.9537469

>>9537464
>Stop projecting.
What do you mean?

>> No.9537481

>>9537466
>Of course it does, otherwise it wouldn't be evidence against B.
It would be, if it was present. A condition has to occur for it to change the probability of an event.

>Rolling an odd number on a die decreases the probability of a 6 irregardless of whether it's "present".
If you don't have the evidence that an odd number was rolled, the probability doesn't change. Even though you know that IF you did have that evidence, it would. You're just mincing words.

>> No.9537486

>>9537481
>It would be, if it was present.
You've started mincing words again, now what is it you mean by "was present"?

>> No.9537488

>>9537481
>If you don't have the evidence that an odd number was rolled, the probability doesn't change.
Correct, that's the absence of evidence, which is why absence of evidence is not evidence of absence, since the probability doesn't change.

>> No.9537489

>>9537481
>Even though you know that IF you did have that evidence, it would.
There's no "having" evidence, P(B|A) can be calculated irregardless of whatever this "having" might mean.

>> No.9537490

>>9537486
"was present" = "A occurred"

>> No.9537494

>>9537488
>Correct, that's the absence of evidence, which is why absence of evidence is not evidence of absence, since the probability doesn't change.
I meant that the probability doesn't decrease.

>> No.9537498

>>9537489
>There's no "having" evidence
Your attempt to deny a common phrasing exists does not convince me.

>P(B|A) can be calculated irregardless of whatever this "having" might mean.
I said it could be calculated. Thanks for agreeing with me.

>> No.9537501

>>9537490
>"was present" = "A occurred"
Define "occurred".

>> No.9537504

>>9537494
>I meant that the probability doesn't decrease.
Correct, since the probability doesn't change, absence of evidence is not evidence of absence.

>> No.9537510

>>9537498
>Your attempt to deny a common phrasing exists does not convince me.
You're simply mincing words again, can you keep things mathematical?

>> No.9537513

>>9537498
>I said it could be calculated.
Right, P(B|A) can be calculated irregardless of whether evidence "is present" (whatever that may mean).

>> No.9537514

>>9537501
I should not have to define basic terms of probability theory. Educate yourself if you don't understand.

https://en.wikipedia.org/wiki/Event_(probability_theory)

>> No.9537517
File: 25 KB, 283x262, Titor_insignia.jpg [View same] [iqdb] [saucenao] [google]
9537517

The evidence of absence meme reminds that once long ago it could be taken for granted that evidence of presence is just what it sounds like.

>> No.9537518

>>9537504
That definition leads to a contradiction. See >>9537129

>> No.9537521

>>9537513
Yes, you can calculate whatever hypothetical you want. So what?

>> No.9537523

>>9537514
Yes, that's the definition of an event. Did you mean to post the definition of "occurred"?

>> No.9537528

>>9537523
I never claimed it was. I shouldn't have to define basic terms. Educate yourself if you don't understand.

https://en.wikipedia.org/wiki/Event_(probability_theory)

>> No.9537535

>>9537528
>I never claimed it was.
Then why respond to me asking for the definition of "occurred" with irrelevant links?

>> No.9537537

>>9537518
>That definition leads to a contradiction. See >>9537129
There's no contradiction in that post. Could you elaborate?

>> No.9537541

>>9537535
I think it's a highly relevant link if you don't understand the basic terms of probability theory. Otherwise, why ask me you define it?

>> No.9537543

>>9537521
>Yes, you can calculate whatever hypothetical you want. So what?
So the probability is updated irregardless of whether you "have" or "have not" evidence (whatever that may mean).

>> No.9537544

>>9537541
>Otherwise, why ask me you define it?
Because you're mincing words and whatever substance of your claims has been offloaded onto a word you refuse to define.

>> No.9537546

>>9537537
The contradiction is that C is both the presence of evidence for B and the presence of evidence against B. You can refer to the prior posts on this so don't ask me about it further unless you have something new to discuss.

>> No.9537547

>>9537546
>The contradiction is that C is both the presence of evidence for B and the presence of evidence against B.
Can you show mathematically how that follows?

>> No.9537548

>>9537541
>I think it's a highly relevant link if you don't understand the basic terms of probability theory
How could it be relevant if it doesn't define the term I asked you to define?

>> No.9537549

>>9537543
>So the probability is updated irregardless of whether you "have" or "have not" evidence (whatever that may mean).
Updated inside of a hypothetical, so what? You're confusing your imagination for the actual application of conditional probability based on all the information that you have.

>> No.9537553

>>9537544
If my argument actually rests on the exceedingly common use of a word in probability theory, I would say that's a good place to have it rest.

>> No.9537554

>>9537549
>Updated inside of a hypothetical, so what?
What part of it is "hypothetical"? Would the calculation change outside of this so-called "hypothetical"?

>> No.9537556

>>9537553
>If my argument actually rests on the exceedingly common use of a word in probability theory
Which is defined as...? You're an expert in probability theory so I expect you should understand the words you're using.

>> No.9537558

>>9537554
>What part of it is "hypothetical"?
The part where you are considering the effect of a condition that has not yet occurred.

>> No.9537560

>>9537558
>The part where you are considering the effect of a condition that has not yet occurred.
You've offloaded your claim onto this mysterious "occurred" again. What part of the calculation changes if whatever has "occurred" or "not occurred"?

>> No.9537564

>>9537560
>What part of the calculation changes if whatever has "occurred" or "not occurred"?
If A occurs then the chance of B increases. That is what P(B|A) > P(B) means. If A does not occur then the chance of B decreases. That is what P(B|~A) < P(B) means.

>> No.9537570

>>9537564
So none of the calculation changes, since if A doesn't occur then P(B|A)>P(B) is still true.

>> No.9537574

>>9537570
P(B|A)>P(B) is not a calculation, it's an inequality.

What's being calculated is the probability of some event based on the current state of information. If the state of information changes, the calculation can change.

>> No.9537581

>>9537574
>P(B|A)>P(B) is not a calculation, it's an inequality.
Mincing words again.

>What's being calculated is the probability of some event based on the current state of information. If the state of information changes, the calculation can change.
What state of information changes the calculation of P(B|A)>P(B)?

>> No.9537584

>>9535327

Problem is using probabilities in the first place, something is either true or not.

Probabilities represent the percentage of an outcome you would get with infinite trials.

>> No.9537586
File: 50 KB, 645x729, 1515194851321.png [View same] [iqdb] [saucenao] [google]
9537586

>>9537584
>Frequentism

>> No.9537598

>>9537564
>If A occurs then the chance of B increases.
You've offloaded your claim onto this mysterious "occurs" again.

>> No.9537618

>>9535327
there's an absence of evidence for anthropomorphic climate change, but that doesn't mean there's evidence of absence for anthropomorphic climate change

>> No.9537635

>>9537618
>there's an absence of evidence for anthropomorphic climate change
There's much more evidence present than absent.

>> No.9537807

>>9535327
This is what happens when you throw around notation without thinking carefully about what it actually means.

>> No.9537830

>>9537807
Yes, only by faulty notation can you claim things like "The absence of your name from the attendance list is evidence that you were absent from class." Clearly such a statement is too ridiculous to be true.

>> No.9537847

>>9535327
>google Kim Oyhus
>"physicist and programmer"
Not surprised that she has such a poor grasp on probability.

>> No.9537856

>>9537830
>The proof is correct because I found a specific example where it holds
No.

>> No.9537858

>>9537830
Just because alien didn't raise their hand when we called their name doesn't mean they don't exist. It just means they either didn't hear you call their name, or weren't there to raise your hand.

Students forget to sign the attendance sheet all the time.

I only mention aliens because I am used to this absence of evidence meme in context with aliens all the time. I haven't read the full thread to see if this is the case this time or not

>> No.9537867

>>9537830
>Yes, only by faulty notation can you claim things like "The absence of your name from the attendance list is evidence that you were absent from class." Clearly such a statement is too ridiculous to be true.
My name is absent from tomorrow's attendance list, is that evidence I'll be absent?

>> No.9537870

>>9537830
But an attendance list is evidence.

>> No.9537884

Non-sequitor.

>> No.9537905

>>9535327

Def 1. Should just be P(B|A) > P(B)

Absence of evidence on it's own is not an event. You need to define an event that can make B less likely (i.e. going out and actually searching for B for an extended period of time), which would also depend on factors of the search unrelated to the probability of B and A.

I am continually not finding evidence of elephants in my bedroom right now, so according to this logic they don't exist.

>> No.9537908

>>9537905
>Absence of evidence on it's own is not an event.
This. If anything it's the empty set, for which P(A|empty set) is not even defined which is why the "proof" in the OP is invalid.

>> No.9537926

>>9537867
BTFO

>> No.9537935

>>9537905
>Def 1. Should just be P(B|A) > P(B)
This doesn't work either, since then applying Bayes Theorem gives
P(A|B)P(B)/P(A) > P(B)
so
P(A|B) > P(A)

so using Definitions 2 and 3 gives us the conclusion:
"absence is evidence of absence of evidence", another meaningless linguistic trick like in the OP

>> No.9538351

>>9537856
Who are you quoting?

>> No.9538356

>>9537858
>Just because alien didn't raise their hand when we called their name doesn't mean they don't exist.
Yes, but it is evidence that they don't. Evidence is not the same thing as proof.

>> No.9538359

>>9537867
Your name being on tomorrow's attendance list is not evidence that you will be in attendance tomorrow, so no. Read the OP carefully.

>> No.9538362

>>9537870
Evidence of what?

>> No.9538389

>>9537905
>Def 1. Should just be P(B|A) > P(B)
That's already equivalent to Def 1.

>Absence of evidence on it's own is not an event.
Sure it is, it's the complement of presence of evidence.

>You need to define an event that can make B less likely (i.e. going out and actually searching for B for an extended period of time),
That event was defined in the OP. And no, you don't need to search for a certain amount of time to satisfy that definition.

>I am continually not finding evidence of elephants in my bedroom right now, so according to this logic they don't exist.
No, according to this logic they are less likely to exist.

>> No.9538391

>>9537908
See >>9536605

>> No.9538965

>>9537935

So? I have no problem with P(A|B) > P(A)

If something exists there is more likely to be evidence for it.

>>9538389

My problem with the OP is defining the complement of A as absence of evidence.

In a way it is, but it's such weak evidence that's pretty much insignificant. Let's take the case of elephants in my room and use Def 1 from OP (P(B|A)>P(B|-A))

We have from Bayes rule:

P(B|-A)=P(-A|B)P(B)/P(-A)

P(-A) is pretty much one, since I can confirm there is no elephant in my room. However, P(-A|B) is nearly zero, since why would and how would an elephant get to my room even if it did exist? For most cases you're pretty much just saying that P(B|A)>0, which tells you nothing.

The probability that an elephant is in my room is greater if they exist, but it's like saying the very low probability that they'd be in my room in the first place makes it so that inequality is always satisfied, giving you no information.

>> No.9538976

>>9538965

P(-A) is pretty much one, since I can confirm there is no elephant in my room. However, P(-A|B) is nearly zero, since why would and how would an elephant get to my room even if it did exist? For most cases you're pretty much just saying that P(B|A)>0, which tells you nothing.

Shit, I meant it's nearly 1, so it cancels with the denominator and you just get P(B) again.

Ignore the rest of that comment.

>> No.9538984

>>9538389
Ah, never-mind. You're right. I'm just being pedantic at this point.

>> No.9539002
File: 1.25 MB, 245x200, 1409185646366.gif [View same] [iqdb] [saucenao] [google]
9539002

>>9535327
I think it's trippy!

>> No.9539008
File: 2.19 MB, 2964x1896, 1514242796218.jpg [View same] [iqdb] [saucenao] [google]
9539008

>>9539002
Fucking almost and without a script.

>> No.9539159

>>9538965
>In a way it is, but it's such weak evidence that's pretty much insignificant.
Its strength is determined by how much we would expect the evidence to be present if B was present. That's it. It could be weak or strong but it's evidence either way.

What information you get out of this is to not ignore the absence of evidence, which humans frequently do. For example if I showed you a frog and told you that only the males of this species croak, most people would ignore the fact that the frog didn't croak while you were observing it and say that the frog has a 50% chance of being male. In reality it can't have a 50% chance of being male if you consider all the information you have, including its lack of croaking, since females never croak while males have a chance of croaking.

>> No.9539189

>>9538965
>So? I have no problem with P(A|B) > P(A)
You have no problem with "absence is evidence of absence of evidence"?

>> No.9539204

>>9538391
>See >>9536605
The first sentence there is a non-sequitur.

>> No.9539309

>>9538965
Just ignore the guy posting above, he's a shitty troll.

>> No.9539321

>>9535327
The definitions just don't really make sense, the first defines evidence as a relation between events and then the second defines "absence of evidence" in a nonsensical way. Definitions 2 and 3 are equally problematic since any event can be written as (not not event), so every event is trivially both "absence of evidence" and "absence".

>> No.9539333

>>9537847
>>google Kim Oyhus
>>"physicist and programmer"
>Not surprised that she has such a poor grasp on probability.
This is probably the most likely source of the issue. This "Bayesian perspective" is more of a weak heuristic than anything rigorously mathematical.

>> No.9539544

>>9539321
>the first defines evidence as a relation between events
It defines when one event is evidence of another.

>then the second defines "absence of evidence" in a nonsensical way.
How so?

>Definitions 2 and 3 are equally problematic since any event can be written as (not not event), so every event is trivially both "absence of evidence" and "absence".
A is defined as evidence for B, so a = ~A is not "every event."

>> No.9539558

>>9539544
>How so?
Because it allows for meaningless conclusions such as in >>9539189

>> No.9539578

>>9535327

absence of evidence does not equal evidence of absence.

for example : i see no evidence OP has a dick but does that really mean he is dickless just because i cant see his tiny little toggler?

>> No.9539579

>>9539558
The absence of B increases the probability of evidence for B being absent. It makes perfect sense.

>> No.9539628

>>9535327
A male hears about this and his first thought is "I should try to understand this claim so that I can verify it myself". A female's first thought is "wait, what if AoE *is* EoA? What if a female can prove all those males wrong? I should try and do that!"

>> No.9539632

>>9539628
>females REEEEEEEEE
>>>/r9k/

>> No.9539635

>>9539632
Have you ever seen a male "copyright" a proof?

>> No.9539641

>>9539635
Kim Oyhus is a male...

>> No.9539644

>>9539641
>Kim Oyhus is a male...
There's an absence of evidence for this claim (note that this isn't evidence of absence).

>> No.9539654

Wouldn't P(B|A) > P(not B|A) make more sense for definition 1?

>> No.9539664

>>9539189

Yes. I changed my mind.

>> No.9539666

>>9539654
Why would that make more sense? You can have evidence for an unlikely event.

>> No.9539679

>>9539666
>You can have evidence for an unlikely event.
Why does my alternative not allow for that?

>> No.9539685

>>9539666
>Why would that make more sense?
Doesn't that fit with the intuition for evidence? That it makes something more likely than the alternative

>> No.9539740

>>9539679
I'll give you an example. I throw a fair die. The chance of it having landed on 1 is 1/6. If you're given the information that the result is odd, that chance is now 1/2. But according to your definition this is not evidence the die landed on 1, dive it did not increase the chance of the result being 1 to greater than 1/2, even though it increased that chance by a lot.

>> No.9539744

>>9539685
Evidence makes something more likely than if there wasn't evidence. You can have weak evidence that doesn't make something more likely than not likely.

>> No.9539829

>>9539654
So then if A has no effect on or decreases P(B) but P(B) is still greater than 1/2, A is automatically evidence? No.

>> No.9539840

>>9536608
>Which is correct since the absence of evidence for one thing is evidence for something else.
and what would that be?

>> No.9539845

>>9539840
See the OP.

>> No.9539862

>>9539744
>Evidence makes something more likely than if there wasn't evidence.
Evidence is just the event in the conditional, it doesn't have to make a given event more likely or not. The definition is for whether it's evidence "of" another event or not.

>> No.9539874

>>9539829
>So then if A has no effect on or decreases P(B) but P(B) is still greater than 1/2, A is automatically evidence?
Any event with non-zero probability can be evidence, in the case you described it would be evidence of B.

>> No.9539879

>>9539862
>Evidence is just the event in the conditional, it doesn't have to make a given event more likely or not. The definition is for whether it's evidence "of" another event or not.
Yes, we're talking about the definition of evidence for B. There is no such thing as evidence irrespective of some other event.

When you say Definition 1 should be P(B|A) > P(not B|A) you are saying that should be the meaning of "A is evidence for B." But this clearly fails since A and B could be independent and P(B|A) > P(not B|A) could still hold.

>> No.9539882

>>9539879
>There is no such thing as evidence irrespective of some other event.
The evidence is just the event that you're updating your probability based off of, as described here
https://en.wikipedia.org/wiki/Bayesian_inference#Formal_explanation

>> No.9539884

>>9539874
>Any event with non-zero probability can be evidence, in the case you described it would be evidence of B.
I don't see how you could call A evidence of B if B decreases the probability of B being true. Evidence should support what it is evidence of.

For example, let's say I am 90% sure that Mary was murdered by her husband. Then I am given new information that tells me there is another suspect for the murder, and that probability drops to 80%. Under your definition, the information that there is another suspect is evidence that Mary is the murderer!

>> No.9539889

>>9539884
>Evidence should support what it is evidence of.
Yes, evidence for an event should support that event more than the complement of that event.

>> No.9539896

>>9539882
>The evidence is just the event that you're updating your probability based off of, as described here
Yes, that is true under the definition in the OP, so I don't see how that's relevant. All evidence updates the probability of some event, either because it is evidence for that event or against it.

Again, you attempted to define "A is evidence for B," not "A is evidence."

>> No.9539902

>>9539889
How does the introduction of a new suspect support the event "Mary is the murderer?"

>> No.9539909

>>9539896
> All evidence updates the probability of some event, either because it is evidence for that event or against it.
Is knowing you rolled an even number evidence for or evidence against the event that you rolled either a 1 or a 2?

>> No.9539916

>>9539902
>How does the introduction of a new suspect support the event "Mary is the murderer?"
Why would it need to?

>> No.9539920

>>9539909
>Is knowing you rolled an even number evidence for or evidence against the event that you rolled either a 1 or a 2?
Neither. I didn't say any event, I said some event.

>> No.9539925

>>9539920
>Neither. I didn't say any event, I said some event.
What event does the evidence {1,2,3,4,5,6} increase or decrease?

>> No.9539926 [DELETED] 

>>9539916
Because P(Mary is the murder | there is a new suspect) = 0.8 > 0.2 = P(Mary's husband is not the murder | there is a new suspect)

>> No.9539927

>>9539925
What probability of an event *

>> No.9539928

>>9539916
Because P(Mary is the murderer | there is a new suspect) = 0.8 > 0.2 = P(Mary is not the murderer | there is a new suspect)

>> No.9539930

>>9539925
>What event does the evidence {1,2,3,4,5,6}
That's not evidence since it doesn't update the probability of some event.

>> No.9539936

>>9539930
>That's not evidence since it doesn't update the probability of some event.
P(E|{1,2,3,4,5,6}) is well-defined for all events E of the toss of a die, it's equal to P(E).

>> No.9539941

>>9539936
I don't see how that responds to my point. Again, evidence supports some event. That is a fundamental necessity of any definition of evidence.

>> No.9539945

>>9539941
>Again, evidence supports some event. That is a fundamental necessity of any definition of evidence.
There's no such notion of "support" in a probability space, an event is just that, an event.

As in https://en.wikipedia.org/wiki/Bayesian_inference#Formal_explanation, you update your probability P(B) to P(B|E) based on evidence E, without any necessary reference to P(B|not E).

>> No.9539952

>>9539928
>Because P(Mary is the murderer | there is a new suspect) = 0.8 > 0.2 = P(Mary is not the murderer | there is a new suspect)
You can simply write this but why is this the case? Can you formalize this? What is P(there is a new suspect)? What is P(Mary is the murderer and there is a new suspect)? In what probability space does the experiment you're modeling allow for outcomes like "Mary is the murderer" and "there is a new suspect" and that "there is a new suspect" only acutely decreases the probability that "Mary is the murderer"?

>> No.9539974

>>9539945
>There's no such notion of "support" in a probability space, an event is just that, an event.
Again, we are not arguing over the definition of evidence, we are arguing over the definition of A is evidence for B. So if there is no notion of support in a probability space, then you have invalidated your own interpretation.

But in fact we do see the notion of support in Bayesian inference:

>That is, if the model were true, the evidence would be more likely than is predicted by the current state of belief.
Which is equivalent to P(B|A) > P(B)

>As in https://en.wikipedia.org/wiki/Bayesian_inference#Formal_explanation, you update your probability P(B) to P(B|E) based on evidence E, without any necessary reference to P(B|not E).
Under that definition there cannot ever be an absence of evidence, since the absence of one conditional is the presence of its complement. Therefore this has nothing to do with the phrase "absence of evidence is evidence of absence."

>> No.9539985

>>9539952
>You can simply write this but why is this the case?
It does not matter why, it's a reasonably realistic example. If you are trying to argue that this cannot occur you need to show me why it cannot. If not, then respond to the argument.

>> No.9539992

>>9539985
>It does not matter why, it's a reasonably realistic example.
Of course it matters why, otherwise it doesn't correspond to anything either mathematical or in reality.

> If you are trying to argue that this cannot occur you need to show me why it cannot.
It's not clear why it can occur. I'm not going to try to argue your own point for you. If all you can provide are inequalities in undefined probability spaces then it's not clear that you have any argument to begin with.

>> No.9539994

>>9539974
>Again, we are not arguing over the definition of evidence, we are arguing over the definition of A is evidence for B.
As in https://en.wikipedia.org/wiki/Conditional_probability#Use_in_inference, we have the usual interpretation:
>The conditioning event is interpreted as evidence for the conditioned event.

>> No.9540005

>>9539974
>>That is, if the model were true, the evidence would be more likely than is predicted by the current state of belief.
>Which is equivalent to P(B|A) > P(B)
And if you read further, we see
>If the belief does not change, P ( E ∣ M ) P ( E ) = 1 ⇒ P ( E ∣ M ) = P ( E )... That is, the evidence is independent of the model.
And so as I said in the example of a die toss, {1,2,3,4,5,6} is in fact evidence.

>> No.9540007

>>9539974
>Under that definition there cannot ever be an absence of evidence, since the absence of one conditional is the presence of its complement.
Why does "the absence of one conditional is the presence of its complement" imply that there cannot be an absence of evidence?

>> No.9540024

>>9539992
>Of course it matters why, otherwise it doesn't correspond to anything either mathematical or in reality.
That doesn't follow, how does it not correspond to mathematics or reality?

M = "Mary is the murderer"
S = "A new suspect is found"

P(M) = 0.9
P(S) = 0.09
P(S|M) = 0.08
P(M|S) = 0.8

Therefore by your definition a new suspect being found is evidence that Mary is the murderer.

>> No.9540026

>>9539994
>The conditioning event is interpreted as evidence for the conditioned event.
See the last part of >>9539974

>> No.9540033

>>9540005
>And so as I said in the example of a die toss, {1,2,3,4,5,6} is in fact evidence.
Then there cannot be an absence of evidence, since any event in a probability space supports, is neutral to, or against another event. As I said, a good definition of evidence requires that the evidence supports something.

>> No.9540041

>>9540033
>Then there cannot be an absence of evidence, since any event in a probability space supports, is neutral to, or against another event.
What is this new "neutral" you've added into your argument? And whatever it is, why does "any event in a probability space supports, is neutral to, or against another event" imply that there cannot be an absence of evidence?

>> No.9540053

>>9540041
>What is this new "neutral" you've added into your argument?
I didn't add it, the article you posted did:
>If the belief does not change... That is, the evidence is independent of the model.

I reject the definition in this article, since it's clearly just using evidence as a word for the conditional without regard to whether it supports the hypothesis or not. The phrase "absence of evidence is evidence of absence" on the other hand is referring to genuine evidence for the hypothesis, P(B|A) > P(B|~A), which is described here:

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.658.238&rep=rep1&type=pdf

>> No.9540067

>>9540041
If evidence for B is simply any conditional that updates B, then the absence of evidence is evidence of absence, since P(absence | absence of evidence) can be calculated as long as P(evidence of absence) > 0, which you just claimed is true.

>> No.9540066

>>9540053
>I reject the definition in this article, since it's clearly just using evidence as a word for the conditional without regard to whether it supports the hypothesis or not. The phrase "absence of evidence is evidence of absence" on the other hand is referring to genuine evidence for the hypothesis
"absence of evidence is evidence of absence" also uses evidence as a word for the conditional without regard to whether it supports the hypothesis or not, which is why it's modified by "of absence".

>> No.9540070

>>9540067
>If evidence for B is simply any conditional that updates B, then the absence of evidence is evidence of absence, since P(absence | absence of evidence) can be calculated as long as P(evidence of absence) > 0, which you just claimed is true.
I assume you meant P(absence of evidence)>0 otherwise what you wrote is not true, and when did I apparently make the absurd claim that P(absence of evidence)>0?

>> No.9540073

>>9540066
>"absence of evidence is evidence of absence" also uses evidence as a word for the conditional without regard to whether it supports the hypothesis or not
Wrong. "absence of evidence" refers to evidence that supports B and "evidence of absence" refers to evidence that supports the absence of B.

>> No.9540075

>>9540070
>I assume you meant P(absence of evidence)>0
Yes.

>when did I apparently make the absurd claim that P(absence of evidence)>0?
You seem to be arguing against my assertion that under your definition, the absence of evidence is impossible. So which is it?

Either we have the absence of evidence is possible, in which case the phrase "the absence of evidence is evidence of absence" is proven true under your definitions. Or we have the absence of evidence is impossible, which means your definitions do not apply to the phrase.

>> No.9540079

>>9540075
>You seem to be arguing against my assertion that under your definition, the absence of evidence is impossible. So which is it?
No, I'm asking where I made the claim that "P(absence of evidence)>0".

>> No.9540085

>>9540079
You didn't make the claim, you seemed to argue against it. Either you've proved exactly what you were arguing against, or you've rendered it irrelevant. Which is it?

>> No.9540100

>>9540085
>You didn't make the claim, you seemed to argue against it.
I don't recall arguing against "P(absence of evidence)>0", can you point to the post(s) you're referring to?

>> No.9540236

Not necessarily true. For example me going outside and not seeing a nuclear holocaust is evidence that a nuclear holocaust didn't happen

>> No.9541046

>>9540100
See >>9540041

To answer your question more succinctly, if all possible events are evidence, then the absence of evidence is an impossible event.

>> No.9541056
File: 271 KB, 3840x2160, TrollFace.jpg [View same] [iqdb] [saucenao] [google]
9541056

>>9535327
Is this the bird of an epic new meme?

>> No.9541071

>>9541046
>if all possible events are evidence
What do you mean by "possible"?

>> No.9541131

>>9541071
P(A) > 0

>> No.9541146

>>9541131
Why such a restrictive notion?

>> No.9541164

>>9541146
What do you mean? According to the Wikipedia article, any conditional which updates the probability of H is evidence. That means P(E) > 0.

>> No.9541612

>>9537847
>>9539333
He's a certified retard and a massive pseud. Here's a link to his site http://kim.oyhus.no/
>tries to refute the existence of God, but only refutes the existence of the Christian God
>unnecessarily uses symbolic logic to do this, like a pseud
Also
>tries to prove that correlation is evidence of causation
>uses Bayesian inference
What a retard lmao