17. Bayesian Statistics
HTML-код
- Опубликовано: 5 сен 2024
- MIT 18.650 Statistics for Applications, Fall 2016
View the complete course: ocw.mit.edu/18-...
Instructor: Philippe Rigollet
In this lecture, Prof. Rigollet talked about Bayesian approach, Bayes rule, posterior distribution, and non-informative priors.
License: Creative Commons BY-NC-SA
More information at ocw.mit.edu/terms
More courses at ocw.mit.edu
I been studying and working with mathematical statistics for almost 10 years. This is the first time I actually understand (intuitively) what the Bayesian approach is
Explain
@@Eizengoldt Explain what...?
@@sobhhi what is the bayesian approach
@@Eizengoldt The guy made a whole 2 hour lecture about it. I don't know if I can so simply collapse it for you into a RUclips comment.
But if I had to: It's introducing "out of data information" to bias the solution of statistical estimation. The hope is that with good information, you can find a better solution than if you only used the data.
39:05 This is it. Weighting definitions are where the magic lies.
Great lecture and a great eraser! Could totally still see after he used that
*Student: Professor, can you write a little bigger so we can see?
*Professor: It might be the board. Let me just change this blackboard to a whiteboard really quick. This eraser should help.
Why cant all professors teach that good? Why did we need to suffer until this lecture?
A great lecture. At 38.58 he should have said Likelihood of observing the data given that the data is distributed according to parameters theta
Exactly thanks !
I updated my knowledge on bayesian statistics :)
Crazy, we estimate the probability of the mind. Bad that I don't already know the language. More homework if too much free time is a good way to progress. I like it. Means we should know where to put our mind to. There never should be too much. There is always more. Thanks...
this fancy new automated camera following the professor gives me a vertigo.
Beta distribution can be bimodal when both a and b are less than 1.
He basically can write with the eraser instead!
😅
when you have to pee with prior belief that you don't know why you put a pie or where it is then the best approach is to squared it with iid. p.s. don't forget to include pmf/pdf.
1:02:00 Gaussian Inference w/Bayesian Methods
Thanks! man
Haven't started.. Thanks already :)
Many Thanks Prof.
as a French, I really appreciate his English accent (too much "you know" tho)
It's not English as in "England".
Would you please explain more, 1:01:06, why expectation not average? for beta(a,b) the expectation is average.
This is at least my third time to watch this video, barely I feel I learned clearly. Thanks Lord!
43:42 MCMC
interesting to see you here..
Very good content to understand Bayesian inference. Thanks prof.
Was anybody else screaming out "LIKELIHOOD" at 38:45??
How would something like a Bayes factor relate to a prior distribution?
difficult to follow explanation an difficult to see 'foggy' blackboard
not the best MIT professor in terms of clarity but still amazing lecture
MIT should invent a better quality borad cleaner. Its hard to read the writngs on the board when it is cleaened.
excellent lecture.thanks a lot
43:20 goes hard as fuck 💯💯💯😤😤👊👌 Now I finally can understand what is going on when i'm using BayesFactor
why does a professor today still waste time writing lectures on greenboards , instead commenting on greenboards the students impressions ?
what is c-cell ?
It is much easier using a discrete random variable X to explain the posterior probability.
One third into the lecture, I'm lost.
Camera man terrible.
Fucking terrible
These videos are helpful, but can't see lecture 16?
Beta distribution may be bimodal when both parameters are less than 1, and the curve is U shaped. Why they don't use whiteboards but these ugly blackboards?
Zdravo!
Can you just stop walking while you are talking because the camera kept panning, it made me feel headache
He looks a lot like Thomas Bayes.
Great lecture. Does anyone have good book tips about bayesian data analysis on longitudinal and time series data structures with R/Jags?
More specifically, I have a case with several subjects and several observations within the subjects. In another case I only have one subject. In both cases, the consecutive observations are supposedly time-related.
Unless you think of probability as belief in bayesian statistics, the whole framework makes no sense.
Statistics aids human epistemology, it's not a replacement for it, even a UMP test result can be rejected as a definite method of learning on other relevant grounds.
joe derosa teaches stats?
Finally，firstly time in this course saw so many cleaner writing XD. Eraser is horrible!!
I know that chalk...
wow ..people really still spending 100k a year to listen to this nonsense.
~21:45 0.52 is zero point five two, not point fifty two.---- basic mathematics.
ab tu math sikhayega
You understood what he meant to the exact precision intended, and so did everyone else. Ridiculous pedantry helps nobody.
Dude he french. zero virgule cinquante deux. he is translated directly.
Omg dude, you're so smart!