Physics Guy wrote: ↑Fri Sep 03, 2021 8:45 pm
Lem wrote: ↑Fri Sep 03, 2021 7:45 pm
Consider what this means. It means that if Joseph Smith did NOT write the Book of Mormon, it is an absolute certainty, with a probability of 1, that it is a type of Early Modern English text. No other options. It can't be a direct hebrew or Reformed Egyptian or mesoamerican translation, or anything else. It MUST be a "filtered and managed" text that is mostly Early Modern English, but written so that 19th century readers understand the Early Modern English. Without exception.
I didn't appreciate this until recently, but this is indeed exactly how the Sharpshooter Fallacy often appears in (bad) Bayesian inference.
My example from a few weeks ago here was my claim to be a prophet based on my producing exactly the long string of random characters that was God's specific message (according to my sect's doctrine).
The sharpshooter declares that the probability of the data, given his hypothesis, is one. That's precisely the Sharpshooter fallacy. The accurate analysis, in contrast, has to be accurate about how likely the data really are, given the hypothesis.
Excellent point, physics guy. I got a kick out of your example you linked to when you wrote it; it was quite interesting to review it again in this new 'episode 9' context.
And in addition to the sharpshooter fallacy, KR continues to not understand the idea of independence. Here's the latest, from this Early Modern English episode:
Summary.
Taken together, the syntactic and semantic evidence of Early Modern English provide two strong, independent lines of evidence against Joseph authoring the Book of Mormon.
The probability of seeing both kinds of evidence by chance would be the product of their respective probabilities, or p = 5.24 x 10-24.
So KR is multiplying the two probabilities, because he considers that observing both syntactic Early Modern English and semantic Early Modern English, in an Early Modern English text, to be INDEPENDENT pieces of evidence.
That would be like arguing that you saw the attached right leg of a two-legged alien in a motion-activated video, and then you ALSO saw the attached LEFT leg of a two-legged alien, right next to the right leg, IN THE SAME VIDEO, and therefore, you conclude that the right leg sighting and the left leg sighting are independent of each other in a two-legged alien, therefore you have two completely INDEPENDENT sightings of a two-legged alien.
Do you sense some irritation? if a student presented this in their group project, it would get an F. That's why I require at least two meetings with me before presentations, so I can help them avoid such egregious errors. (and Fs.)
How could the statistician who reviewed Rasmussen's work, on behalf of the Interpreter, let something like this go? I'm getting the sense that the statistical review was about the same quality as the Interpreter's usual playground peer review.