Bayes in Science and Everyday Life: Crash Course Statistics #25

Today we're going to finish up our discussion of Bayesian inference by showing you how we can it be used for continuous data sets and be applied both in science and everyday life. From A/B testing of websites and getting a better understanding of psychological disorders to helping with language translation and purchase recommendations Bayes statistics really are being used everywhere!
Will Kurt's A/B Testing Example: www.countbayes...
Crash Course is on Patreon! You can support us directly by signing up at / crashcourse
Thanks to the following Patrons for their generous monthly contributions that help keep Crash Course free for everyone forever:
Mark Brouwer, Erika & Alexa Saur, Glenn Elliott, Justin Zingsheim, Jessica Wode, Eric Prestemon, Kathrin Benoit, Tom Trval, Jason Saslow, Nathan Taylor, Divonne Holmes à Court. Brian Thomas Gossett, Khaled El Shalakany, Indika Siriwardena, SR Foxley, Sam Ferguson, Yasenia Cruz, Eric Koslow, Caleb Weeks, Tim Curwick, D.A. Noe, Shawn Arnold, Ruth Perez, Malcolm Callis, Ken Penttinen, Advait Shinde, William McGraw, Andrei Krishkevich, Rachel Bright, Mayumi Maeda, Kathy & Tim Philip, Jirat, Eric Kitchen, Ian Dundore, Chris Peters
--
Want to find Crash Course elsewhere on the internet?
Facebook - / youtubecrashcourse
Twitter - / thecrashcourse
Tumblr - / thecrashcourse
Support Crash Course on Patreon: / crashcourse
CC Kids: / crashcoursekids

[@sinecurve9999]

P("watched video" | "wrote 'First!' in the comments") = P("wrote 'First!' in the comments" | "watched video") * P("watched video") / P("wrote 'First!' in the comments")

[@mattkuhn6634]

Bayesian statistics are the foundation of Statistical Machine Translation systems. These days, however, SMT is usually outperformed by systems that use Recurrent Neural Networks, which don't really use Bayes' theorem or markov models Loving the series so far!

[@uzomaohajekwe7149]

Do neural networks have anything to do with statistics?

[@reader1236]

I love the channel but I'm afraid I don't get this. The beauty blogger could just compare the number of clicks with/without the picture and then keep adding the pictures - where is the need for "prior beliefs" in this? Also, it seems that the better the evidence, the more weight it gets, so in essence we're back to frequentism, aren't we?

[@mw79863s]

With no label on graph axes at 6:19 , something difficult to understand is made worse.

[@tutukkunoor]

True, I can't figure what is being plotted. Since there is only one email each sent to each group, what does Y Axis mean is not clear o.O

[@tutukkunoor]

I'm assuming maybe many emails were sent, and moving ahead

[@ellsbells1235]

I think the x axis is click rate (in terms of a probability, so 0.3 for example refers to 30% of people receiving the email clicking it) and y is the relative probability of the hypothesis represented on the graph being true (I.e. the likelihood that the click rate is the true click rate) :)

[@kiane9465]

Crash Course Architecture please!

[@Eirikr430428]

Didn't know I wanted this until now!

[@hntddt1]

Metal gear solid easter egg found

[@YuzuruA]

What I love about bayesian statistics is that, if it existed in the times of Galileu, the pope could simply say that the evidence would not be enough to change his prior beliefs.

[@innocentoctave]

Generally speaking, organized, hierarchical religions don't like uncertainty: it leaves room for dissent.

[@bangladesh6027]

Generally speaking, organised , hierarchical religions reject all evidence that conflicts with their prior beliefs instead of incorporating it

[@rafaelricardovilorio602]

This course on statistics is so good! I've enjoyed a lot learning.

[@Wiggles42]

Thanks for the video! Very nice. Would you consider making a video about Bayesian inverse problems? i.e. using Bayes as an alternative to least-squares curve fitting problems?

[@mesplin3]

Wiggles42 I think it's the same.

[@XrollhaX]

Can't you use this to figure out what a person thinks of something? Like, can't we give the other values and find P(H1)?

[@lolalucxyz]

Arnon Tillmann You can figure out what probability someone used as a prior given their weighting of the evidence and their result, but for that to work the person in question must have used Bayes theorem to come to their conclusions. People aren't perfect Bayesian machines, so in most cases they won't deal with evidence correctly, and they will probably fail to express their priors as proportionate probabilities. What you can do is figure out an approximation of what you or somebody else should think of something, if they were rational.

[@andreasnautsch5575]

Thanks for the first steps to Bayes factors. Question: why are Bayesian methods are put to confidence tests as the natural choice would be credibility tests (i.e., keeping consistency within a Bayesian framework, whereas confidence tests would undermine fundamental Bayesian principles, i.e., the model-ability of randomness in terms of probabilities)? // sorry for pettifogging - cheers & thanks for the video!

[@صفاءعلاءابوعلي]

Hhhh I study now statistic Good video

[@jacobluna305]

I got a B+ in college statistics, so you can call me a real life statistician if you want. LOL

[@Commando303X]

"Malingerer," not "maligner."

[@manuelsputnik]

wooo data science baby

[@kebman]

Markov chain gang. Yeah, they're sentenced to life hard labour.

[@centori2011]

I'm a fan of Bayes' theorem but I fail to understand what she's attempting to explain.

[@CompilingInput]

Fred Rogers for novice the explanation is way too complicated hate that

[@centori2011]

CompilingInput Not that it's too complicated, I find that she's focused on the more complex and least useful aspects of Bayes' theorem for an introductory lesson. I find Bayes' very useful and don't think you need a computer to use it. If she wants to get people to use it I don't thinks she going about it in the right way.

[@CompilingInput]

Fred Rogers true, for novice description can be in more layman terms then slowly cud be advanced into more complex applications, we climb one step a time not 5 or more 😁😁

[@lolalucxyz]

Well, this isn't a life hacks channel. If you want an explanation that is designed to maximise utility in real life while minimising effort/exposure to details, you should check out Julia Galef's video.

[@centori2011]

Lucca Fehn I wouldn't watch crash course if it was a life hack, no where does it imply that. "Crash course" as the name implies, that's what I expect like a lesson as in school but on KZbin instead. One area where I like to apply Bayes' is evaluating the probabilities of historical events or something practical from our every day lives not "whether house elves have lower intelligence than wizards" to quote her. What the hell is she talking about? She's focusing on the complexities, I don't see how that's practical to anyone but an advanced user. You can still use Bayes' and exclude both those variables she discussed. Bayes' is used to solve conditional probabilities and this is the formula P(H∣E)=P(E∣H)P(E)P(H) it just describes the probability of an event, that's why it's useful.

[@TheNael92]

They have "the theory that will never die" book in the background. Glad that this channel does their research on the subject.

[@valor36az]

What a fantastic channel why are there no recent posts

[@locotx215]

7:50 - Unexpected Mr. Robot

[@tutukkunoor]

Difficult Lecture :( Couldn't understand the emails and DID example properly.

[@Gumikrukon]

Thank you!! Great stuff :)

[@Johnjo85]

first!!

[@ArawnOfAnnwn]

TIL D.I.D. patients are able to control which personality they're under...

[@aNytmare]

I don't think it is that easy. D.I.D. is more complicated than just asking a patient to switch. But we the video is not about D.I.D , but about Bayes

[@nothisispatrick4644]

The moment you showed me math formulas I just averted my gaze from the screen of my phone. Because it brings back bad memories

[@Pfhorrest]

bay-EE-shun!

[@remdog37]

Rhubarb

[@abdullaalzarooni2837]

11th

[@kontonkitsune2670]

Lol I’m a nerd. “Crash course released a new vid” YES! *clicks*

[@tutukkunoor]

At 2:30, why does she say all probabilities of getting tails are equally likely? Aren't probabilities of getting all or no tails in a fair coin in multiple flips less than probabilities of some heads and some tails? What is the meaning of that flat line?

[@MadaxeMunkeee]

It’s a subtle point, but the line represents your belief that the coin is fair. If you take theta to be the odds the coin gives tails, then if the coin is fair theta = 0.5. If the coin only gives tails then theta = 1 and if heads only is possible then theta = 0. Because you don’t have any prior information about the coin, it’s reasonable to model your belief in the value of theta with what would be called a flat (or uniform) prior. This distribution gets updated as you observe more coin tosses and gain information about theta.

[@ajaxhopper9859]

crash course, stoooop with these titles, it hurts

[@JuliusUnique]

Why is is called "Bayes" tho, when in reality it's just empiciral stochastic?

[@aNytmare]

Thomas Bayes was a priest, who talked about Math/statistics in his sermons... pretty cool huh?

[@JuliusUnique]

not really, I think calling it empirical stochastic would save much time for many people. For example I'd have known what this video is about and could have saved 11min of my life

[@zoikles1]

No, in reality it's just Bayes Theorem. It's called that because Thomas Bayes was the one who first proved the theorem.

[@MadaxeMunkeee]

What about 'The Empirical Stochastic Probability Theory of the late Reverend Thomas Bayes'?

[@will2see]

You talk too much.