Proof of Immortality III

[John Jones]

,
Dave,
- I'll take a look, but even if it's all covered, it isn't well organized

Nothing you have discussed has ever been. In fact, your efforts seem to be dedicated to making every discussion disorganized and opaque.

I don't see that changing after all these years.

 

[JayUtah]

- I'm trying to lay out all your 'reservations' in an organized fashon...

You can "organize" the debate all you want. You're an expert at avoiding debate by "organizing" it for years at a time without ever managing to address its glaring faults.

1) Is the formula I'm using appropriate for determining the posterior probability of OOFLam?
- Dr Hoerl said that it is...

Here is your report of his latest meeting.

js,
- Good.
- I'm waiting for something (brief) in writing from Dr.Hoerl.
- He doesn't agree with me -- so far. But also, so far, I don't understand his reservation...
- We still have a lot of discussion to do, but he has a publication due in two weeks, and won't be able to continue till then. I'll post his (brief) reservation in writing as soon as I get it.

He does not agree with you. And you are so far unwilling to discuss his reservations with him. As with "Sally" we never seem to get anything from Dr. Hoerl that doesn't come through you, and doesn't end up sounding like it came from you. I don't believe you have been conversing with a professional statistician.

No, it is not the right formula. And, in fact, to phrase it in terms of "the right formula" is rather provincial. It is a line a hack Hollywood screenwriter would put in the mouth of a professor. It is the correct formula for computing a posterior probability, but this is not a posterior probability problem. You don't understand how Bayesian analysis works. And I mean you don't understand it at a fundamental level. You commit the same conceptual errors as every other fringe theorist does who thinks he can use it to transform invention into fact. To put those misconceptions into the supposed mouth of an expert is just compounding your dishonesty.

You seem to believe you can maintain this bluff indefinitely. You don't seem to realize that there are people here who do understand how these models work and can tell just how badly you're faking it. You were given several essays describing your errors, which you ignored. You ignored them because you can't respond intelligently to them. So you've now invoked a deus ex machina to absolve you of the need to discuss in detail your errors.

You propose to put the question of your competence in statistical reasoning beyond the scope of this inquiry. It most certainly is not beyond the scope, and I do not intend to allow you to escape that accountability.

 

[JayUtah]

Lets see if we can list the various reservations to using Bayes in this context.

I already laid them out, months ago, in one of your other threads, at considerable length. If you weren't willing to address them now, why would you be willing to do so again? How many times must you be presented with your errors before you will stop "organizing" them and start addressing them?

ETA: And let's be clear about the burden of proof here. You're the one saying you can prove your case mathematically, and you're the one claiming that Bayesian analysis is the proper mathematics to do that. You are on the hook to justify that claim of applicability. Suggesting that your critics have the onus to affirmatively disqualify it is shifting the burden of proof. The "reservation" against using Bayes here is simply the disputation of your claim that it applies.

Substantiate that claim.

 

[zooterkin]

- I'm trying to lay out all your 'reservations' in an organized fashon, and address them one at a time. Maybe, you guys will ultimately win this debate -- but, if we can actually do this in an organized way, we'll have a 'book' to which we can direct our friends, and where they can decide for themselves. Immortality or Bust?

1) Is the formula I'm using appropriate for determining the posterior probability of OOFLam?
- Dr Hoerl said that it is -- so hopefully, we can skip that sub-issue and move on to #2?

I think the significant word there is 'posterior'.

 

[Jabba]

Agatha,
- For whatever reasons, Dr Hoerl has never gotten back to me after our meeting. I have to suspect that I'm just bothering him, so I hesitate to contact him again...
- I'll see if I can get someone at SUNY to talk to me.

- If you can remind me of one of your reasons for disputing the use of Bayesian statistics here, I'll see if I can answer it. Otherwise, I'll follow Dave's advice and go back to the beginning of the beginning thread and try to itemize your (plural) reservations in the first few pages.

 

[JayUtah]

For whatever reasons, Dr Hoerl has never gotten back to me after our meeting.

So then the last word on the subject from him is what I reproduced above, in which you report that he does not agree with you. Yet today you have said he confirmed your use of Bayes as the proper model. It seems you can't keep straight from month to month what your consulted experts have said.

I'll see if I can get someone at SUNY to talk to me.

See if you can tell us, based on your own knowledge, why you think Bayes is an appropriate model for proving this -- or any -- conclusion. Tell us in your own words. You've spent years on this debate, so you can certainly spare a day to spell out in a suitably lengthy original essay your rationale.

This practice of simply reporting the alleged endorsements of experts you say you've consulted is tantamount to anonymous expertise. It doesn't work with "Sally" and it won't work elsewhere. If you propose to use sophisticated math to solve a problem, you don't get to defer to some other person and place a discussion of your competence to invoke the sophistication. You seem to think you can trump detailed rebuttals of your claims with some second-hand anonymous imprimatur.

If you can remind me of one of your reasons for disputing the use of Bayesian statistics here, I'll see if I can answer it.

Asked and answered. I told you several times I do not repeat myself to interlocutors who brag about how little of the debate they actually read. If you now realize you missed something through your deliberate dereliction, then your penance is to review the thread in its entirety until you're caught up.

Otherwise, I'll follow Dave's advice and go back to the beginning of the beginning thread and try to itemize your (plural) reservations in the first few pages.

Do not "itemize" them. Do not "organize" them. Address them.

 

[ferd burfle]

- I'm trying to lay out all your 'reservations' in an organized fashon, and address them one at a time.

No, What you're "trying" to do is foist on us yet another fringe reset. We know that and you know that.

1) Is the formula I'm using appropriate for determining the posterior probability of OOFLam?
- Dr Hoerl said that it is -- so hopefully, we can skip that sub-issue and move on to #2?

Nice try. Please tell us what Dr. Hoerl said said about how you are applying the formula. Using the correct formula is necessary but not sufficient. I'll bet he told you just what Jay has told you, that your inputs can't be numbers plucked from a nether orifice.

 

[godless dave]

Otherwise, I'll follow Dave's advice and go back to the beginning of the beginning thread and try to itemize your (plural) reservations in the first few pages.

That wasn't my advice.

 

[Jabba]

So then the last word on the subject from him is what I reproduced above, in which you report that he does not agree with you. Yet today you have said he confirmed your use of Bayes as the proper model. It seems you can't keep straight from month to month what your consulted experts have said...

Jay,
- Dr Hoerl disagreed with my P(E|H), but not my use of the specific Bayesian formula. He said that he would send me a brief summary of his reservation, and we could get back together -- but, he never did send the summary.
- I have just contacted SUNYA by email...

 

[jond]

Do not "itemize" them. Do not "organize" them. Address them.

This, Jabba! THIS!!!

 

[theprestige]

Jay,
- Dr Hoerl disagreed with my P(E|H), but not my use of the specific Bayesian formula. He said that he would send me a brief summary of his reservation, and we could get back together -- but, he never did send the summary.
- I have just contacted SUNYA by email...

Dr Hoerl doesn't matter because he's not here to speak for himself, and you can't be trusted to represent him accurately.

Unless SUNYA comes here to speak for themselves, they don't matter either.

It's all part of yet another one of your fringe resets.

 

[JayUtah]

No, What you're "trying" to do is foist on us yet another fringe reset.

It's the next logical step after punting with the Bukowski and Russell quotes. Whenever he gets painted into a corner he throws those out there to imply that anyone who has drawn a conclusion is a fool. Every Jabba thread involves him declaring that the debate is incurably muddy -- through his equivocal obfuscations, obviously -- and then insisting we go back to first principles to "correct" his mud-stirring.

I'll bet he told you just what Jay has told you, that your inputs can't be numbers plucked from a nether orifice.

I'm sure that's what he did, if in fact there was any such meeting. One of the guiding principles of Bayesian analysis is that the priors are facts, not estimates. Sure, you can put estimates in there and run Bayesian models to see what would happen. But as you can imagine, the results only hold insofar as the estimate holds. You most certainly can't shove whatever you want in there and pretend anything comes out of it that reflects the real world.

The notion of "H given E" and "E given H" as operators in the Bayesian world simply do not work backwards as Jabba proposes to use them. He wants to start with posteriors and work backwards to determine what the prior "must" be. This has been explained to him countless times. "Itemizing" it or "organizing" it again does nothing to address that fatal flaw. But if he uses those weasel words, he can pretend that he's engaging in the debate without actually having to address the lack of merit in his argument.

Bayesian analysis is inherently an iterative process. Which is to say, its strength in generating knowledge lies in its ability to adjust parametric distributions of probability based on new data, in ways that classical statistics can only model with random variables. The iteration is the key. Here, in this invocation, there is no iteration. There is simply the proposition. Jabba has grossly co-opted the language and intent of Bayesian methodology to propose what he insists divides the prior and posterior conditions in this case. But his proposition is not an informational event; it's just more of his numerical proctology.

So on the one hand Jabba simply invents numbers that, in Bayesian analysis, must be measured. And on the other hand, Jabba tries to shoehorn his static proof into the dynamic nature of Bayesian methods by cobbling up the semblance of an event. There is no "event" in Jabba's proof; just two sets of made-up numbers.

I cannot stress enough that when I say Jabba's misunderstanding of Bayes is fundamental, it is fundamental. His notion of its foundational principles isn't even in the right county, let alone the right ballpark.

 

[JayUtah]

Dr Hoerl disagreed with my P(E|H), but not my use of the specific Bayesian formula.

I don't believe you. I really and truly do not believe you. I don't believe for one single minute that any professional statistician or mathematician has even looked at your model here, much less endorsed any part of it.

The notion that a professional would disagree only with some tiny quibble of an input, and not your glaring foundational ignorance, is astounding. You are not even close to an accurate understanding of statistical modeling using any methodology.

And no, I don't care what new imprimatur you've solicited from some expert we'll never get to address ourselves. I want you to address the flaws in your model, not just declare that some person somewhere has wholesale endorsed it.

I've outlined some of your egregious faults above and below. Please address them in detail, in no fewer than 5,000 of your own words.

 

[JayUtah]

That wasn't my advice.

But it's what he desperately needs your advice to have been, so that he can spend the rest of June "organizing" the debate instead of participating in it, and pretending that this is what his critics have asked of him.

 

[gypsyjackson]

If Jabba will answer (unless he already has) whether it is Roger Hoerl or Arthur Hoerl, it would be possible to contact Dr Hoerl as both have published their email addresses.

Not that I'm advocating winding up a professional.

 

[ComfySlippers]

...
He said that he would send me a brief summary of his reservation, and we could get back together -- but, he never did send the summary.
...

Take the hint.

 

[JayUtah]

Not that I'm advocating winding up a professional.

I am occasionally contacted by strangers to ask whether I really did say one thing or another. I generally don't mind, as long as it's a proper requests and it's clear I'm not necessarily amenable to being drawn into whatever debate has invoked it. But being asked to set the factual record straight is not objectionable to me. It's not a wind-up if the question simply asks to confirm what was alleged about him. If anything, Jabba's repeated invocations of him in this thread would be objectionable. To cite an expert, ostensibly in his official capacity, in public implies that Jabba has obtained permission for that use of the expert's knowledge and judgment, and shouldn't object to third parties contacting the expert for verification.

 

[JayUtah]

If Jabba will answer (unless he already has) whether it is Roger Hoerl...

It's Roger. Early in the thread, Jabba reproduced what he claimed was an email from Dr. Hoerl in which he is addressed as "Roger."

 

[JayUtah]

I think the significant word there is 'posterior'.

By which I assume you mean "numerical proctology," as I described it above.

It is by no means wrong to use estimates as priors in Bayesian analysis. It depends on what you want the analysis to show. Using a spectrum of estimates as priors is a common method of constructing decision trees to account for contingencies. However, this does nothing to validate the priors or justify a substitution for fact if fact is available. It simply shows what would be justifiable decisions from new facts and uncertain circumstances. If the priors are considered a distribution of estimated likelihood, as would be proper, then the result of iterating the model over given facts would be a new distribution -- not a single number.

Similarly, one may certainly encode a belief as a quantitative likelihood. This is a common method of constructing decision trees that mix belief and fact. However, the posterior probability distribution function -- even combined with fact that is later observed -- has absolutely no power to confirm the prior belief as fact. Bayesian analysis is not an invertible function -- this inescapable mathematical fact is the basis of the Prosecutor's Fallacy. The results of Bayesian analysis in this mode hold only insofar as the belief. It essentially places the belief beyond the realm of computation and shows only what would be justified, given new information, without questioning the belief. It has absolutely no value for someone who does not share the belief.

Neither of these cases constitutes in any way a proof of anything. The only way Bayesian analysis can be used as a proof is where the priors and parameters are measured facts. And the proof, in the form of posterior probability distributions, is only of a probability. In no way, under any circumstances, by any means, can a probabilistic argument be proof of an allegation of fact. That is, at its very heart, the sum total of Jabba's error.

 

[Loss Leader]

Dr Hoerl doesn't matter because he's not here to speak for himself, and you can't be trusted to represent him accurately.

Unless SUNYA comes here to speak for themselves, they don't matter either.

It's all part of yet another one of your fringe resets.

Hooray for SUNY Albany, my alma mater. Go whatever we were!