Bayesian Statistics Demystified
Вставка
- Опубліковано 6 тра 2024
- The Michael Shermer Show # 429
What's the difference between "conventional" and Bayesian statistics?
Bestselling author and twice-recipient of the Royal Statistical Society’s award for statistical excellence in journalism Tom Chivers helps us delve into the concept of Bayesian reasoning.
If you’ve heard the term before and wondered how it applies to the real world, this one is for you.
SUPPORT THE PODCAST
If you enjoy the podcast, please show your support by making a $5 or $10 monthly donation.
www.skeptic.com/donate/
#michaelshermer
#skeptic
Listen to The Michael Shermer Show or subscribe directly on UA-cam, Apple Podcasts, Spotify, Amazon Music, and Google Podcasts.
www.skeptic.com/michael-sherm... - Наука та технологія
Richard Carrier gave a really thorough, book length, understandable explanation of Bayes's Theorem, "Proving History."
😢😢wwwwe😢😢😢
21:53 exactly. most medical professionals get it wrong most of the time.
therefore, most of our taxes are thrown in the bin and burnt to carbon.
58:00 - Identity protective cognition is the idea that people cling to whatever opinion enhances the glory of their tribe and their status within it. The social cost of giving up some beliefs is often greater than a real desire to seek what is most likely true.
27:03 I also think that was a good way of describing the value of new information but from the opposite direction that most people would take. Obviously in that context he meant that each new condition/constraint we put on an event decreases the overall probability of that event as compared to the same event without that condition/constraint. In his example he used North Korea attacking Hawaii just in general vs North Korea attacking Hawaii when there were some specific santactions and other constraints, obviously the more general version with less constraints is more probable. That same example can sort of show the value of each new piece of information we obtain because it can narrow probability distributions. Said another way" "is it more or less likely that North Korea would attack Hawaii given some specific sanctions are in place or given that they are not in place?". The reason, or rather a reason, that the event with more constraints is less probable than the event with less constraints is that we must also consider the probability of that constraint and then the probability of the event. In his example we would need to first consider the probability that the US would place those constraints on North Korea and then the probability that they would attack. However, given we know the constraint is true or false then it could have a similar narrowing effect, assuming the information has value. As an example, I could say "what is the probability that North Korea will attack Hawaii given I had tacos last night" vs "What is the probability that North Korea will attack Hawaii given the US placed new heavy sanctions on North Korea"
Ahh, talking about the quirks of course because that’s exciting... Setting limits (like “18 years”) is not weird; it’s done every day in process control through guard-banding.
👍
Good episode apart from the superficial part on vaccinations.
it was never mysterious .... just hard 😅
Wasn't it Kierkegaard who said that if God were to speak to you directly, how would you as an individual know if it was really God and not Satan?
It's weird how proponents of this view don't differentiate between bayesian statistics and bayesian epistemology. It's like they think it's the same thing.
Excellent point. They are often conflated.
31:50
there are dozens of people alive on earth today that have cured cancer and are immune to viruses.
admittedly, there arent many of us but... we can prove 99.5% of humanity as erroneous.
personally, i figured out how to digest and process all undesirable pathogens at 23yrs of age. im 44 now and pathology free for two decades.
the only thing that gives me a headache is too many beers the night before.
There are so many "eyewitness reports" about uap's I'm surprised that nobody has sorted them out and performed some kind of analysis. Locations, differences, similarities, etc...
When that's the only data you have it would make sense to do this. Sure are a lot of people who say they are interested, why hasn't someone gone ahead and done it?
You could at least determine the validity of some reports compared to others, and I suspect, much more.
The subject itself isn't that interesting to me but the lack of research aimed at the reports themselves strikes me as very odd.
Personally, my concept of "alien" is something that's never been reported or even imagined. If the aliens landed tomorrow I suspect that exactly zero previous reports will have been confirmed. So, you could build a scale to make comparisons of 'more alien' vs 'less alien' and other priors in the absence of hard evidence. We already have lots of research about psychology and some research on "eyewitness testimony" as it applies to the court of law, but this recent hype in my opinion is an "eyewitness testimony phenomenon".
I don't have the resources or the experience to do this or I would.
Why hasn't someone done this?
Are they REALLY so curious?
Doesn't seem like it.
I believe Jaques Valle has done that ! in his books, he does not reveal what the conclusions of that analysis are.
Too many ads! I guess UA-cam is just pushing people to pay for UA-cam premium.
Maybe start posting this podcast on other platforms?
COVID statistics literacy? Beg to differ. Throughout 2020 everyone discussed only ABSOLUTE infection numbers in United States when relevant data is only relative to population size.
Everyone? epidemiologists?
"..everyone discussed absolute..etc"
this is a false statement, all of the medical and military scammers used relative stats to promote quaxxines. Indeed, most biological science papers use completely erroneous techniques such as - covariance, regression analysis, p-values, normal distributions, averaging and significance tests.
all are false.
a complete bayesian analysis is the only way science can achieve validity.
science is bayes.
simples.
philosophers are centuries ahead of the low grade pursuits of chemistry and biology because of a day to day understanding of inductive logic.
therefore, ninety nine percent of scientists are scammers.
The issue with Baysian reasoning is that it is filled w underlying assumptions that can be 100% wrong. In other words: it is only valid as a first attempt at a general look into a topic. And it can only be used in simple constructs with known variables.
For instance: Climate Change. An a priori impossible subject for bayesian reasoning. In fact: bayesian reasoning is constantly used by climate alarmists to make their flawed data stick. But if the underlying assumptions about the mainstream climate are wrong, baysian reasoning doesnt matter.
Could you give your top three of assumptions about the climate that could be wrong?
Do you include climate scientists with the climate alarmists?
32:09 no, wakefield was vindicated and the case was thrown out.
shermer spreading false information.
countermeasures cause autism.
here's a thought:
i am an exceptionally fit and healthy fellow. i have been such all my life. i travelled a lot during youth.
my parents would often jab me by procedure.
during my youth, i can remember these random pains in my heart. very rare pains. maybe one or two pulses of heart pain/ heartburn once or twice a year, maybe once in two years.
the pain would stop me in my tracks then dissipate.
clearly, i was a lucky one.
now i know the cause of those mysterious heart pains.
so, by bayesian reasoning i can infer that most people who take jabs will experience far worse outcomes than myself. youd have to be a pro athelete or similar to equate a similar baseline.
so, i can confidently bet that the same heart pain will be experienced by all jab users - even the fittest will suffer. (or suddenly drop dead in the fauci sense).
to the athletes: that weird, random, unexpected heart burn/pain....now you know the cause.
fascinating subject, but mediocre guest.
If Michael Shermer did a solo episode on bayes, it might have been more interesting