Interpretering Bayesian Analysis

The catch-all forum for general topics and debates. Minimal moderation. Rated PG to PG-13.
Philo Sofee
God
Posts: 5017
Joined: Thu Oct 29, 2020 1:18 am

Re: Interpretering Bayesian Analysis

Post by Philo Sofee »

I can't wait to see part 4 on the DNA stuff. I've read over 500 pages of scholarly materials from Mormons, non-Mormons, scientists, geneticists, etc. all weekend since friday night. I am exhausted, but gettin a small view of the issues... This is going to be interesting to say the least.

And since Kyzabee mentioned his part 4 was the DNA stuff, the probability of my having jumped on it and reading like a bat outta hell was 1, surely! :D
hauslern
1st Counselor
Posts: 474
Joined: Tue Dec 08, 2020 2:36 am

Re: Interpretering Bayesian Analysis

Post by hauslern »

I am eagerly waiting for how how he deals with the Book of Abraham.
Philo Sofee
God
Posts: 5017
Joined: Thu Oct 29, 2020 1:18 am

Re: Interpretering Bayesian Analysis

Post by Philo Sofee »

hauslern wrote:
Mon Jul 26, 2021 4:33 am
I am eagerly waiting for how how he deals with the Book of Abraham.
Yeah, I saw that was on the agenda. It's gonna wipe him completely out regardless of what he does with the Book of Mormon. I sincerely do hope he realizes that...
Analytics
Elder
Posts: 350
Joined: Wed Oct 28, 2020 3:11 pm

Re: Interpretering Bayesian Analysis

Post by Analytics »

Gadianton wrote:
Sat Jul 24, 2021 2:46 pm
I totally agree, the "evidence" -- at least the stuff so far -- has nothing to do with the Book of Mormon being ancient. The exercise is pointless on its face.
Exactly. Apologists typically argue that the reason mainstream scholars aren't convinced by their arguments is because they aren't familiar with them. They argue that Michael Coe, for example, is speaking out of ignorance when he says the Book of Mormon has nothing to do with actual Mesoamerican history. If only he would have seriously read what John Sorenson et. al. have written on the topic, then he'd be qualified to opine on the topic. They accuse him of being closedminded for not taking Mormon apologetics seriously.

So I'm trying to imagine how an openminded mainstream scholar would react to this. Imagine. Rasmussen gets on the agenda at a serious conference for Mesoamerican anthropology. He announces he has a breakthrough discovery--he's discovered an accurate translation of an authentic ancient Mesoamerican manuscript. It is over 800 pages long, and details their history for over 1,000 years. This is, perhaps, the most important New World archeological discovery of all time. Rasmussen explains that while the original manuscript is gone, we can be certain that the translation is accurate and authentic with a confidence of over 99.9999999999999999999999999999%.

The scholars in the conference are sitting on the edge of their seats, ecstatic to be participating in this momentous occasion. Rasmussen begins to lay out his proof to the manuscript's authenticity to this audience.

1- It is over 800 pages long, and thus couldn't have been written in the 19th Century.
2- It was translated by somebody who says he saw God and gave conflicting accounts of the story.
3- It was dictated in just 65 days.....

Rasmussen explains to the scholars that with the help of modern statistical analysis, the likelihood ratio of all three of things happening with an authentic Mesoamerican document rather than 9th century fiction are higher than a million to one.

Being openminded, the scholars in the conference are really impressed and can't wait to hear more.
User avatar
Gadianton
God
Posts: 3843
Joined: Sun Oct 25, 2020 11:56 pm
Location: Elsewhere

Re: Interpretering Bayesian Analysis

Post by Gadianton »

Analytics wrote:This is, perhaps, the most important New World archeological discovery of all time. Rasmussen explains that while the original manuscript is gone, we can be certain that the translation is accurate and authentic with a confidence of over 99.9999999999999999999999999999%.
Right, we'd have greater confidence that the Book of Mormon is an ancient record of real Nephites, then the once-lost city of Troy getting dug up in modern times really is the city of Troy. The day Zarahemla is dug up along with all the surrounding cities, with transliterated names matching exactly the Book of Mormon names, people will shrug their shoulders. "Duh, what part of the Book of Mormon being 800 pages, produced in 65 days, and containing 15th century English didn't you understand?" Save yourselves the cost of the shovels next time.
User avatar
Dr Moore
Endowed Chair of Historical Innovation
Posts: 1812
Joined: Mon Oct 26, 2020 2:16 pm
Location: Cassius University

Re: Interpretering Bayesian Analysis

Post by Dr Moore »

Meet Richard Carrier, the authority to whom KR appeals when pressed on the scientific validity of his process.

https://en.wikipedia.org/wiki/Richard_Carrier
Carrier's methodology and conclusions in this field have proven controversial and unconvincing to specialists,[3][4][5] and he and his theories are often identified as fringe.[6][7]


His process has generated substantial criticism by experts. I guess by Mopologetic standards, that means "all I do is win, win, win no matter what!"

Here's just one from the Wikipedia page, which says everything that ever needs to be said about deploying Bayesian tools against religio-historical claims.
Reviewing On the Historicity of Jesus, Daniel N. Gullotta says that Carrier has provided a "rigorous and thorough academic treatise that will no doubt be held up as the standard by which the Jesus Myth theory can be measured"; but he finds Carrier's arguments "problematic and unpersuasive", his use of Bayesian probabilities "unnecessarily complicated and uninviting", and he criticizes Carrier's "lack of evidence, strained readings and troublesome assumptions."[5] Furthermore, he observed that using Bayes theorem in history seems useless, or at least unreliable, since it leads to absurd and contradictory results such as Carrier using it to come up with low probability for the existence of Jesus and scholar Richard Swinburne using it to come up with high probability that Jesus actually resurrected.[89]


In other words, Bayes provides just enough fog to obscure the thumbs on the scale.
Fence Sitter
High Priest
Posts: 398
Joined: Wed Oct 28, 2020 2:02 am

Re: Interpretering Bayesian Analysis

Post by Fence Sitter »

Very interesting thread.

From someone who does not understand it, would anyone like to explain how using Bayesian Analysis on historical events is different from the Texas Sharpshooter Fallacy?
Analytics
Elder
Posts: 350
Joined: Wed Oct 28, 2020 3:11 pm

Re: Interpretering Bayesian Analysis

Post by Analytics »

Dr Moore wrote:
Mon Jul 26, 2021 6:44 pm
In other words, Bayes provides just enough fog to hide the thumbs on the scale.
I have mixed feelings about this. Consider the following passage from Bart Ehrman's excellent book, Misquoting Jesus:
Bart Ehrman wrote: Textual critics who consider themselves rational eclecticists choose from a range of readings based on a number of pieces of evidence. In addition to the external evidence provided by the manuscripts, two kinds of internal evidence are typically used. The first involves what are called intrinsic probabilitiesprobabilities based on what the author of the text was himself most likely to have written. We are able to study, of course, the writing style, the vocabulary, and the theology of an author. When two or more variant readings are preserved among our manuscripts, and one of them uses words or stylistic features otherwise not found in that author’s work, or if it represents a point of view that is at variance with what the author otherwise embraces, then it is unlikely that that is what the author wrote—especially if another attested reading coincides perfectly well with the author’s writing elsewhere.[statistical concepts emphasized]
On the one hand, if an analysis uses the language of statistics, then it is logically coherent to use the tools of statistics to aggregate and understand the probabilities. In theory, if somebody does this it makes their argument more explicit--it allows the reader to understand how the pieces of the argument fit together and reductionally evaluate it and respond to it.

That is clearly what Richard Carrier was trying to do. In his own words:
Richard Carrier wrote:I intend this book not to end but to begin a debate about this, regarding both its methods and its conclusions. Hence, if readers object even to employing Bayes’s Theorem in this case (or in any), then I ask them to propose alternative models for structuring the debate. If, instead, readers accept my Bayesian approach, but object to my method of assigning prior probabilities, then I ask them to argue for an alternative method of assigning prior probabilities (e.g. if my choice of reference class is faulty, then I ask you to argue why it is, and to argue for an alternative). On the other hand, if readers accept my method of assigning prior probabilities, but object to my estimates of consequent probability, then I ask them to argue for alternative consequent probabilities—not just assert some, but actually argue for them. Because the mythicist case hinges on the claim that these things cannot reasonably be done. It is time that claim was properly put to the test. And finally, of course, if readers object to my categories and sub-categories of evidence or believe there are others that should be included or distinguished, then I ask them to argue the case....

But it is the method I want my fellow historians to correct, replace or perfect above all else. We can’t simply rely on intuition or gut instinct when deciding what really did happen or who really did exist, since that simply leans on unexamined assumptions and relies on impressions and instincts that are often not reliable guides to the truth. We need to make explicit why we believe what we do rather than something else, and we need this as much in history as in any other field.
I completely agree with Carrier here. In theory.

Valid Bayesian reasoning is more tricky than it would seem at first blush, and aside from those of us who really loved story problems in middle school algebra and have a knack for understanding and evaluating the implicit assumptions in a mathematical model, it's easy to think the reasoning is stronger because it is in a Bayesian format rather than simply being more explicit.

In aggregate, is the Bayesian approach worthwhile? Probably not.
Analytics
Elder
Posts: 350
Joined: Wed Oct 28, 2020 3:11 pm

Re: Interpretering Bayesian Analysis

Post by Analytics »

Fence Sitter wrote:
Mon Jul 26, 2021 7:45 pm
Very interesting thread.

From someone who does not understand it, would anyone like to explain how using Bayesian Analysis on historical events is different from the Texas Sharpshooter Fallacy?
I would say that in theory, Bayesian analysis encourages you to ask the right questions, makes your arguments more explicit, and if you know what you are doing, weight them correctly.

Whether you are a Texas sharpshooter depends upon your assumptions. When Kyler says there is a 0.0001% chance that a fictional book would be written in 65 days but a 100% chance that a historical book would be, the second half of that is the Texas sharpshooting. If somebody else were to reply that the chances are 0.00001% for a fictional book to be written in 65 days and the chances are 0.00001% for a historical book to be written in 65 days, he is not Texas sharpshooting.

Texas sharpshooting is an example of GIGO. Bayesian analysis might cause your eyes to glaze over so you can't see GIGO, but it doesn't cause it.
User avatar
Dr Moore
Endowed Chair of Historical Innovation
Posts: 1812
Joined: Mon Oct 26, 2020 2:16 pm
Location: Cassius University

Re: Interpretering Bayesian Analysis

Post by Dr Moore »

Analytics wrote:
Mon Jul 26, 2021 8:32 pm
Bart Ehrman wrote: Textual critics who consider themselves rational eclecticists choose from a range of readings based on a number of pieces of evidence. In addition to the external evidence provided by the manuscripts, two kinds of internal evidence are typically used. The first involves what are called intrinsic probabilitiesprobabilities based on what the author of the text was himself most likely to have written. We are able to study, of course, the writing style, the vocabulary, and the theology of an author. When two or more variant readings are preserved among our manuscripts, and one of them uses words or stylistic features otherwise not found in that author’s work, or if it represents a point of view that is at variance with what the author otherwise embraces, then it is unlikely that that is what the author wrote—especially if another attested reading coincides perfectly well with the author’s writing elsewhere.[statistical concepts emphasized]
This is great if you're comparing 2 manuscripts and need a tool to determine which is more likely to be authentic. Very different from looking at a single manuscript and wanting to deploy Bayes to assess whether an author who claims a divine source is being truthful. In that case, the alt prior might be 100% because who in their right mind claims divine authorship for something but changes nothing about their writing style? And given that correlation between the alt prior and the process behind it, all other "odd" behavior of the text are going to be very highly correlated with this one instance. Meaning, it's irresponsible to try and claim statistical independence among other textual attributes.
Analytics wrote:
Carrier wrote:We can’t simply rely on intuition or gut instinct when deciding what really did happen or who really did exist, since that simply leans on unexamined assumptions and relies on impressions and instincts that are often not reliable guides to the truth. We need to make explicit why we believe what we do rather than something else, and we need this as much in history as in any other field.
I completely agree with Carrier here. In theory.

Valid Bayesian reasoning is more tricky than it would seem at first blush, and aside from those of us who really loved story problems in middle school algebra and have a knack for understanding and evaluating the implicit assumptions in a mathematical model, it's easy to think the reasoning is stronger because it is in a Bayesian format rather than simply being more explicit.

In aggregate, is the Bayesian approach worthwhile? Probably not.
Very well said. Of course, the Dales' paper eloquently laid out the perils of cherry picking, and yet their paper is a masterwork of cherry picking. Carrier cautioning that a Bayesian framework makes it possible to overstate the strength of an argument applies to him as well. For all we know, he had his thumbs on the scale too, possibly without realizing it.

One thing about statistics, which also applies to machine learning and artificial intelligence: more data = better result. Therefore, it is also true that less data = worse result. And we have to recognize that not all data is equal, not all data that orbits a question is relevant to that particular question, most historical data is subjective, and the plural of anecdote is not data.
Last edited by Dr Moore on Mon Jul 26, 2021 9:43 pm, edited 1 time in total.
Post Reply