Well, I'm pretty sure you heard some of them before... But the question is: "are the statisticians I know still alive or they passed away?" 😂😂 I knew Cox from Box-Cox transformations and Rao from Rao-Blackwell and Cramer-Rao, but didn't have a clue about when they lived, so such a surprise they lived till a couple years ago
Well, Nobel died in 1896 and the prize started in 1901, before Von Neumann and Turing were even born, so I'm pretty confident nobody told Nobel that Computer Science existed lol
@@very-normal i just thought that these people deserve recognition. thats the least we can do using the free software we've been using. 🙂 nice videos by the way. i love ur content. always looking forward to your uploads.
Hey, this is an amazing video! Cheers to these great statisticians. Rao taught one of lecturers in undergrad. He could never stop speaking so highly of him!
You can make a video about the biggest unsolved questions in statisticslike the millenium prizes. Determined on the importance of the questions, the difficulty of the question and how statistical they are in their essence. :) My guess would be Andrew Gelman for the 2025 medal since Social Science are among the big 3 of statistics: Physic statistics, Bio-statistics and Social Statistics.He already have a lot of medals from his contributions on causal inference. He has mostly focused on social science with regards to voting patterns but social science is used in many high-tech companies for social medias.
I'll definitely be interested to see who this year's prize goes to. In my opinion Andrew Gelman is definitely in the running. But given how new this prize is, there are others who ought to be considered first.
The bootstrap and Crémer-Rao lower Bound are most important invention in stats in last century - they deserve the recognitions without doubt. My predition: Nobel Prixe of stats for 2025 is James-Stein Estimator resp. their proofs - that was huge surprise for many statisticians and showed that MLE is not the sufficient estimator and contradict to Crémer-Rao lower bound.
A topic thats fascinated me for a long time is the statistics of persuasion. How strong does the evidence need to be to persuade people one way or another? Of course, rhetoric is the main way we persuade other people, but it's a nice thought experiment and a very bayesian challenge
Great video! I'm very curious how the CR bound interacts with the infamous bias-variance tradeoff. I wish you had time to go deeper on the finer points of some of these breakthroughs, but I guess that's the nature of a 'best of' compilation like this :)
The CR bound exists for unbiased estimators, so it gives you a theoretical bound for how good an unbiased estimator can be. For a fixed bias (i.e. none), anything meeting it will have the lowest variance / best efficiency. But people have since realized that allowing for a little bit of bias can create valuable estimators due to an outsized drop in variance. I think LASSO and ridge regression are the most famous examples of this.
Thank you for the videos. The story I heard as a student was Nobel's wife was having an affair with a Mathematician, which is why there is no Nobel Math Prize.
Man do I wish you made these videos when I was doing my bachelors in statistics, would've removed a lot of confusion. Still though I really enjoy watching your channel and I hope your goal of making statistics fun for everyone succeeds!
Nice video! This was very interesting. What about Henderson's linear mixed model equations? It's been used everywhere. The thing is he already died, unfortunately.
Yeah! I’ve been cooking up an MCMC type of video for some time now. Jackknife would be cool too, tho it’s been overshadowed by the bootstrap I feel. Could be a part of a bigger video!
Very nicely presented, I learned a lot and really enjoyed the reasonable pace at which you walked the viewer through the contributions as well as their significance.
@@TheThreatenedSwanMostly yes but the contributions of Paul Cohen, Terrence Tao, Martin Hairer improved software verification and algorithms, medical imaging and climate and financial modelling respectively
"I don't think this is true in general. At some level, it's certainly not true if we're talking about the CRLB of unbiased estimators, because the MLE is sometimes biased. For example, in a uniform distribution on [0,theta], the MLE is biased, and the Fisher Information is not even defined. My guess is that this applies for some "location families", which the normal, binomial, poisson would all be. For a "scale family" like the exponential distribution, in the parameterization where the mean is 1/lambda, I do not believe the MLE meets the CRLB." I quote one of my statistics teachers here. So i am confuse d now- is mle estimators always meet crlb?
My understanding is that the MLE is asymptotically unbiased and efficient. It can still be the case that the MLE itself will be biased, but this bias will go away as the sample size goes to infinity; likewise it’s variance will also approach the CRLB
Has any statistician come up with a statistical function that predicts, with any certainty, their chances of winning the International Prize in Statistics.
I agree with your prediction about Vladimir Vapnik. He would be a worthy recipient. It would also recognise the long term efforts of the Russian probability school.
WAIT! I found out on Wikipedia that there has been a "Wilks Memorial Award" since the sixties! Famous names I know who won the prize are C.R.Rao, Neyman, Cochran, Snedecor and many others... No Idea of it is reserved only to residents in the US though
A question: i(θ) isn't just an approximation of the variance of the MLE based on asymptotical results, and moreover MLEs are very often biased because of Jensen inequality or other reasons, so there could be either unbiased as/more efficient estimators or more efficient biased estimators than the MLE. Am I wrong? I also saw a video about James Stein estimator for example, which doesn't take the MLE to get more efficient *Edit: my broken screen and my poor sight prevented me from seeing the bottom note
Honestly I think Rubin is the best prediction now. I think more statisticians and general researchers would be familiar with his name compared to Vapnik
Actally the economics prize is not a Nobel. It's officially the Sveriges Riksbank Prize in Economic Sciences and was not part of the will of Alfred Nobel Just so the economist could sneak their beak in. Curiously enough The Nobel Foundation threatened legal action for a proposed "Michael Nobel Energy Award" " To the Nobel Foundation the 'Dr. Michael Nobel Award' represents a clear misuse of the reputation and goodwill of the Nobel Prize and the associations of integrity and eminence that has been created over time and through the efforts of the Nobel Committees"
0:15 and already the first blatant mistake. There is no "Nobel Price", i.e. price funded by Alfred Nobel, for economics. The "Nobel price in Economics" is the Imperial Bank of Sweden price for economics commemorating Alfred Nobel. "Even Peace", however, IS a real Nobel Price, as Nobel thought he had created a weapon so potent wars would no longer be possible (i.e. what in reality are nuclear weapons).
@@very-normal nah but seriously though, at least make a shorts with how other prizes are distributed and with some data crunching make statistical predictions especially since you havent done much of those
@@very-normal Also in my textbook, in some questions they use root (n) for t-test and in some places its root (n-1). Standard of error is the root of (variance per statistical individual). There wasnt an explanation as to why root of n-1 is used in some places. lmk asap pls, I have a test on 5th in inferential statistics.
In general, the one using root(n-1) is more correct than root(n) because it makes the estimator unbiased. I put root(n) here because that’s what you get with the MLE for estimating the variance of normally distributed data.
@@very-normal how does a root of (n-1) make a significant difference? A hypothesis test especially in your sampling sizes is gonna be large. diff between root of n and n-1 is gonna be in the 0.000x probably. Also how does it make it unbiased?! from an undergrad of Aswath Damodaran, my understanding was that bias is an error from human judgement. How can it be reduced if not eliminated by subtracting 1? Im highlighting my ignorance rn, but the days of mean median and mode were far more comprehensible.... I am stuck with the simplest of t-tests 😭😭
RA Fisher gets credit for popularizing it, but there were a bunch of people before him who made references to it. There’s a paper called “The Epic Story of Maximum Likelihood” by Stephen Stigler that answers your question more thoroughly
It’s less biology and more so trying to translate biomedical ideas into statistical models But yes, dealing with doctors on statistics has an element of masochism to it sometimes
Bold of you to assume I know the name of any statistician.
Bernoulli has a distribution if i remember correctly 🤔
I would be cauchyous with that one, too.
Well, I'm pretty sure you heard some of them before... But the question is: "are the statisticians I know still alive or they passed away?" 😂😂
I knew Cox from Box-Cox transformations and Rao from Rao-Blackwell and Cramer-Rao, but didn't have a clue about when they lived, so such a surprise they lived till a couple years ago
Sealy of you to think I'm only a disinterested Student
Pearlson ? Bermoulli ? That old guy called binomial ?
To paraphrase Chappelle Roan, C. R. Rao is your favorite statistician's favorite statistician
International Prize in Statistics? IPISS sounds like a proper nickname
starting a petition to make that the official name
The Economics prize was added later. It is not an official one, which is why it says in honor of Alfred Noble. Which is why Math maybe added.
It's funny how people mention it's not official to deride the winners having beliefs they dislike when the peace and literature prizes exist
@@TheThreatenedSwan true
Well, Nobel died in 1896 and the prize started in 1901, before Von Neumann and Turing were even born, so I'm pretty confident nobody told Nobel that Computer Science existed lol
lol that’s fair I’ll give him a pass for that
I heard his wife cheated on him with a mathematician, but google quickly told me that it wasn't true.
@@kodierergHe was never married so that probably didn't happen 😅
Well Charles Babbage and Ada Lovelace already did some amazing work by that time so he could've heard about it
The Zuse prize
i think Ross Ihaka and Robert Gentleman, the designers of R, deserve this prize as well as many students and statisticians use R.
That’s a good one, I didn’t even think about the programming route when I was coming up with my own prediction
@@very-normal i just thought that these people deserve recognition. thats the least we can do using the free software we've been using. 🙂
nice videos by the way. i love ur content. always looking forward to your uploads.
0:53 Yup, that’s me. You may wonder how I ended up in this situation…
Hey, this is an amazing video! Cheers to these great statisticians. Rao taught one of lecturers in undergrad. He could never stop speaking so highly of him!
You can make a video about the biggest unsolved questions in statisticslike the millenium prizes. Determined on the importance of the questions, the difficulty of the question and how statistical they are in their essence. :) My guess would be Andrew Gelman for the 2025 medal since Social Science are among the big 3 of statistics: Physic statistics, Bio-statistics and Social Statistics.He already have a lot of medals from his contributions on causal inference. He has mostly focused on social science with regards to voting patterns but social science is used in many high-tech companies for social medias.
I'll definitely be interested to see who this year's prize goes to. In my opinion Andrew Gelman is definitely in the running. But given how new this prize is, there are others who ought to be considered first.
Exponential distribution entered the room
poor guy won’t remember he did
The bootstrap and Crémer-Rao lower Bound are most important invention in stats in last century - they deserve the recognitions without doubt.
My predition: Nobel Prixe of stats for 2025 is James-Stein Estimator resp. their proofs - that was huge surprise for many statisticians and showed that MLE is not the sufficient estimator and contradict to Crémer-Rao lower bound.
The prize could be given to Willard D. James since he is alive while Charles Stein isn't. :)
Thanks to know that. James and Stein both deserve the prize for their work.
A topic thats fascinated me for a long time is the statistics of persuasion. How strong does the evidence need to be to persuade people one way or another?
Of course, rhetoric is the main way we persuade other people, but it's a nice thought experiment and a very bayesian challenge
Money ,money, money !
Great video! I'm very curious how the CR bound interacts with the infamous bias-variance tradeoff. I wish you had time to go deeper on the finer points of some of these breakthroughs, but I guess that's the nature of a 'best of' compilation like this :)
The CR bound exists for unbiased estimators, so it gives you a theoretical bound for how good an unbiased estimator can be. For a fixed bias (i.e. none), anything meeting it will have the lowest variance / best efficiency.
But people have since realized that allowing for a little bit of bias can create valuable estimators due to an outsized drop in variance. I think LASSO and ridge regression are the most famous examples of this.
Thank you for the videos. The story I heard as a student was Nobel's wife was having an affair with a Mathematician, which is why there is no Nobel Math Prize.
Man do I wish you made these videos when I was doing my bachelors in statistics, would've removed a lot of confusion. Still though I really enjoy watching your channel and I hope your goal of making statistics fun for everyone succeeds!
It would be great to see a video on active inference and how it relates to Bayesian statistics
BRO, thank you for this channel and your work! Truly truly insightful!
Nice video! This was very interesting. What about Henderson's linear mixed model equations? It's been used everywhere. The thing is he already died, unfortunately.
Can we get a video on the Jackknife method or on MCMC?
Yeah! I’ve been cooking up an MCMC type of video for some time now. Jackknife would be cool too, tho it’s been overshadowed by the bootstrap I feel. Could be a part of a bigger video!
@@very-normal MCMC is used in lattice QCD and quantum gravity. I'd be interested to see in what other fields they're used in.
Thank you Christian. Love all your videos. Thank you for making them, I'm learning a lot
Very nicely presented, I learned a lot and really enjoyed the reasonable pace at which you walked the viewer through the contributions as well as their significance.
Remember, data is only random from a frequentist perspective. Data is fixed according to Bayesian statistics!
Isnt it disturbing that the fields medal only gives the winner 15000$? I mean math is the base of our infrastructure
Doesn't the vast majority not have real applications?
@@TheThreatenedSwanMostly yes but the contributions of Paul Cohen, Terrence Tao, Martin Hairer improved software verification and algorithms, medical imaging and climate and financial modelling respectively
There is not a nobel prize in economics.
it is an award the swedish central bank hands out.
The problem with max likelihood, is that it leads to overtraining
Statistics is the workhorse for the sciences.
"I don't think this is true in general. At some level, it's certainly not true if we're talking about the CRLB of unbiased estimators, because the MLE is sometimes biased. For example, in a uniform distribution on [0,theta], the MLE is biased, and the Fisher Information is not even defined. My guess is that this applies for some "location families", which the normal, binomial, poisson would all be. For a "scale family" like the exponential distribution, in the parameterization where the mean is 1/lambda, I do not believe the MLE meets the CRLB."
I quote one of my statistics teachers here. So i am confuse d now- is mle estimators always meet crlb?
My understanding is that the MLE is asymptotically unbiased and efficient. It can still be the case that the MLE itself will be biased, but this bias will go away as the sample size goes to infinity; likewise it’s variance will also approach the CRLB
Has any statistician come up with a statistical function that predicts, with any certainty, their chances of winning the International Prize in Statistics.
I agree with your prediction about Vladimir Vapnik. He would be a worthy recipient. It would also recognise the long term efforts of the Russian probability school.
You get hierarchical modeling and the variance of estimates (almost) for free with Bayesian analysis. Take the Bayes pill and make a video about it.
ya boi is fully pilled up, a hierarchical model video would be a good one
WAIT! I found out on Wikipedia that there has been a "Wilks Memorial Award" since the sixties! Famous names I know who won the prize are C.R.Rao, Neyman, Cochran, Snedecor and many others...
No Idea of it is reserved only to residents in the US though
Actually, I thought about talking about this award and the COPPS Presidents award, but it got removed in the editing process 😅
Great channel. Good luck and thanks for the videos
Where's the Galton prize? Or at least one after Pearson
A question: i(θ) isn't just an approximation of the variance of the MLE based on asymptotical results, and moreover MLEs are very often biased because of Jensen inequality or other reasons, so there could be either unbiased as/more efficient estimators or more efficient biased estimators than the MLE.
Am I wrong? I also saw a video about James Stein estimator for example, which doesn't take the MLE to get more efficient
*Edit: my broken screen and my poor sight prevented me from seeing the bottom note
great video! I didn't know about the price & i'm doing a master in stats haha
Professor Vapnik absolutely deserves this prize 😁I had him in mind from the beginning of the video!
My guess would be Donald Rubin, known for his work in propensity score and EM algorithms.
Honestly I think Rubin is the best prediction now. I think more statisticians and general researchers would be familiar with his name compared to Vapnik
Loved the content! Beautifully explained !!
Isn't Cox's work kind of an extension of GLMs with a particularly useful GLM?
Awesome video as always!
5 categories, economics is named after the two novel prize
Actally the economics prize is not a Nobel. It's officially the Sveriges Riksbank Prize in Economic Sciences and was not part of the will of Alfred Nobel
Just so the economist could sneak their beak in.
Curiously enough The Nobel Foundation threatened legal action for a proposed "Michael Nobel Energy Award"
" To the Nobel Foundation the 'Dr. Michael Nobel Award' represents a clear misuse of the reputation and goodwill of the Nobel Prize and the associations of integrity and eminence that has been created over time and through the efforts of the Nobel Committees"
There's no time to go over survival statistics? Well I'm doomed.
it is sad that fields medal gives only 15k
Well thank god no one told them CS exists or else we'd have an arbitrary prize for the easiest form of applied math.
Great video
Nobel had a wife. She had a lover. He was a mathematician.
So, no Nobel Prize for mathematics or mathematicians.
Google told me this wasn’t true
Well ... Google _might_ be right.
0:15 and already the first blatant mistake. There is no "Nobel Price", i.e. price funded by Alfred Nobel, for economics. The "Nobel price in Economics" is the Imperial Bank of Sweden price for economics commemorating Alfred Nobel. "Even Peace", however, IS a real Nobel Price, as Nobel thought he had created a weapon so potent wars would no longer be possible (i.e. what in reality are nuclear weapons).
cool
That was super interesting, thank you!
Well, I knew Florence Nightingale, but I was pretty sure she was not the one to win this 😄
lol have you read The Lady Tasting Tea by David Salsburg by chance
use statistics for predicting the winner
🧠
@@very-normal nah but seriously though, at least make a shorts with how other prizes are distributed and with some data crunching make statistical predictions especially since you havent done much of those
@@very-normal Also in my textbook, in some questions they use root (n) for t-test and in some places its root (n-1). Standard of error is the root of (variance per statistical individual). There wasnt an explanation as to why root of n-1 is used in some places. lmk asap pls, I have a test on 5th in inferential statistics.
In general, the one using root(n-1) is more correct than root(n) because it makes the estimator unbiased. I put root(n) here because that’s what you get with the MLE for estimating the variance of normally distributed data.
@@very-normal how does a root of (n-1) make a significant difference? A hypothesis test especially in your sampling sizes is gonna be large. diff between root of n and n-1 is gonna be in the 0.000x probably. Also how does it make it unbiased?! from an undergrad of Aswath Damodaran, my understanding was that bias is an error from human judgement. How can it be reduced if not eliminated by subtracting 1? Im highlighting my ignorance rn, but the days of mean median and mode were far more comprehensible.... I am stuck with the simplest of t-tests 😭😭
Who said, "all models are wrong, but some are useful"?
I think it’s usually attributed to statistician George Box
What’s the background music
I looked up “calm music” on Storyblocks and took a track that I liked
what is the difference between biostatistics and biostatics?
biostatistics is applying statistics to biological contexts, biostatics is when I can’t pronounce the former correctly
Thinking Judeah Pearl or Donald Rubin?
Pearl won the Turing Award.
There should be absolutely no award for economics whatsoever, what a fudged up "field"
Who did invent MLE?)
RA Fisher gets credit for popularizing it, but there were a bunch of people before him who made references to it.
There’s a paper called “The Epic Story of Maximum Likelihood” by Stephen Stigler that answers your question more thoroughly
Judea Pearl for the 2025 prize?
solid guess! My causal inference guess was Donald Rubin, but I stuck with my ML guess
I expect Pearl wouldn't be nominated because he already won the Turing Award.
C.R rao prolly my fav statistician
You really should research your stories. Nobel intentionally omitted mathematics because a mathematics scoundrel stole his wife.
Lol the irony of this statement
nobel never had a wife as he never got married
"Biostatistics"
Oh god, there's a biology degree with even more statistics? That's straight up masochism.
It’s less biology and more so trying to translate biomedical ideas into statistical models
But yes, dealing with doctors on statistics has an element of masochism to it sometimes
I think we need to talk Christian. If there’s away where I can talk to you privately, I would love to talk to you.
no thank you
Isnt it disturbing that the fields medal only gives the winner 15000$? I mean math is the base of our infrastructure