We'll be discussing the Bayesian statistics part.
What assurances do we have that you'll present your
entire model, not just the parts of it you think your authority will endorse? What assurances do we have that you'll faithfully and accurately represent his criticism? What assurances do we have that you're even in fact in contact with him and aren't just dropping his name?
By now you must realize I'm quite familiar with various common patterns of fringe argumentation. The offline consultation is fraught with them. To deal with our last concern first, very often fringe claimants will assert some relationship to some well-known or easily discovered authority who -- it is represented -- endorses the claimants beliefs. But great lengths are typically taken to prevent the claimant's critics from verifying that supposed association.
If the authority actually exists and actually is consulted, the claimant almost always presents to him a sanitized, emasculated, motte-and-bailey version of the controversy. I remember one guy who asked a physicist to confirm his computations of radioisotopic decay. The problem with his claim was that the phenomenon he was discussing was not governed by that model. So while the expert confirmed his working of the model, the claimant had conveniently forgotten to mention what problem he was trying to solve with it.
Based on the thirty or so claimants who have mentioned offline authority and the all-but-one who invoked one and misrepresented the claim, I estimate the probability that you will misrepresent your present claims to Dr. Hoerl at p > 0.967.
Then there's the answer. If the authority exists and is actually consulted and actually gives an answer, about half the time the claimant cherry-picks the answer for only the parts that favor his beliefs.
Now who would do such a thing? Exactly you. You have a documented history of presenting weakened facsimiles of your arguments in order to garner agreement. You have a documented history of appealing to people like "Sally" whom you hide from your critics. You have a documented history of misstating and misusing other people's quotes unfairly.
Now please address these concerns.
Being quite generous (IMO), I estimated P(OFL) to be .99.
Your "generosity" is irrelevant. If you propose to quantify something, you must show how the quantity was arrived at. p = 0.99 may seem like a very high probability to you, but I work in a world dominated by component reliability requirements along the lines of p = 0.99999 -- five "nines." And I get to see those pass through a series of combinatorics -- including Bayesian models -- that rapidly bring those down to a system reliability of around p = 0.8.
One of the classic examples of a Bayesian system involves medical tests that have a measured reliability of 0.99. When you apply the
real Bayes model (not your silly cobbled-up travesty of it), some very intriguing posterior probabilities emerge. That's why Bayes is so powerful -- as with all good statistical models, it challenges our intuition.
But it's also why Bayes is so often misused by fringe theorists. It alludes to quantified beliefs, which leads them wrongly to conclude it can inject rigor into what is never more than simply a statement of belief. Bayes can accurately show you the effect of information upon belief, but it does not test the truthfulness either of the information or the belief.
I suspect that you really want to know how I estimated P(H).
Naturally. You have a habit of telling us you'll explain where all these "estimates" come from, but never doing it.
But more importantly we really want to know whether Dr. Hoerl endorses the part of your model where you just make up the numbers you put into it. As I have explained many times to you, the strength of Bayesian reasoning is no greater than the validity of the priors. If you simply invent them, however "generously," then there simply is no strength.
