This explanation is excellent. It is crystal clear to explain why is the inverse relationship between variance and second derivative, and why is second derivation, and plus why it is negative! Bravo, Prof.Ben!
This was my intuition when studying ML estimators in statistics, but never got a straight answer about it from my teachers. Happy to see others think of it through a geometric lens! Great video
Studying for actuarial exams and the material just throws Fisher Information at you with no context. This will help me understand exactly what we are expected to do in the calculations. Thank you
Really appreciate videos like this where the aim is to provide an intuitive explanation of the concepts as opposed to going into detail on the maths behind them. Thanks.
Hi Mr. Lambert, I just want to take a moment to thank you for taking the time to make these videos on UA-cam. They are very easy to understand and by watching your videos I have been able to understand my statistical theory and bayesian statistics courses more as an undergrad. Thanks a lot and I wish you all the best!
This video makes me very clear about one thing, that I find it strange how hard it obviously is for professors to provide some clear intuition. Why must it be so hard to be pedagogical when you really know something, which I expect a professor does. This is a working day of headache over horrible handouts made understandable in 5 mins.
OK 3 months ago, I thought I understood this video. After I learned more statistic. Now I understand what is going on. I didn't quite understand the concept 3 months ago.
you said we add the negative sign, because the second derivative is negative after a certain value, and the negative sign is added to correct for that negative. what about when the second derivative is positive? doesn't the negative sign make the second derivative negative then? of what use will that be?
Thank you for this video. I have watched this video many times over the years. The simplicity, intuition, visuals, clarity, and ease, are nothing less than brilliant. It has always helped whenever things get fuzzy. Just a small request or a question if you may: Calling vertical axis "likelihood of the data" makes it a bit confusing! Instead, should it not be "likelihood of the parameter" that is L( theta; data). And this "likelihood of the parameter" then happens to be equivalent to f(data|theta)? So, y axis should not be called L(data|theta)?
Hi ben, thank you so much for your videos, i am studying quantitative ecology and do not have a strong mathematical background - your lessons really help! May I ask how the different values of theta are generated (along the x axis)? I assume the MLE expression stays constant and that the parameter estimates vary due to sample variation but in my case I only have one sample. I am a bit confused whether variance of the MLE is actually referring to variance in the parameter estimate due to sampling error. Secondly, in order to calculate the variance, must the 2nd derivative be evaluated for the value of theta which gives the MLE? I hope these questions make sense!
i dont know hat he is using but I sometimes use app.liveboard.online/ . It also allows you to chose different backgrounds for a board and different colors and to livestream your drawing from your tablet/smartphone to PC which i often use as it is better to draw by hand/pen then by mouse.
THIS MAKES SO MUCH SENSE!! Thank you so much for explaining this more clearly in a few minutes than my textbook could do in a few hours!
In 7m and 59s you explained it better and more clearly than many 2h university lectures combined.
Like everyone else said, very well explained. I feel way less jittery about this whole entire concept. Thank you in 2019!
This explanation is excellent. It is crystal clear to explain why is the inverse relationship between variance and second derivative, and why is second derivation, and plus why it is negative! Bravo, Prof.Ben!
This was my intuition when studying ML estimators in statistics, but never got a straight answer about it from my teachers. Happy to see others think of it through a geometric lens! Great video
Studying for actuarial exams and the material just throws Fisher Information at you with no context. This will help me understand exactly what we are expected to do in the calculations. Thank you
Really appreciate videos like this where the aim is to provide an intuitive explanation of the concepts as opposed to going into detail on the maths behind them. Thanks.
I think this small video worth few 2hrs lectures in a university.
lmao true
I just watched a 1 hour lecture about Cramer-Rao Lower Bound and you are totally right :P this was waaay more informative.
Damn. You explained this so well. I never have any idea what my professor is talking about, but videos like this help SO MUCH. Thank you!
Hi Mr. Lambert, I just want to take a moment to thank you for taking the time to make these videos on UA-cam. They are very easy to understand and by watching your videos I have been able to understand my statistical theory and bayesian statistics courses more as an undergrad. Thanks a lot and I wish you all the best!
This video makes me very clear about one thing, that I find it strange how hard it obviously is for professors to provide some clear intuition. Why must it be so hard to be pedagogical when you really know something, which I expect a professor does. This is a working day of headache over horrible handouts made understandable in 5 mins.
This was a fantastic intuitive explanation - thank you!
Beautifully explained my friend- intuition is almost always as important as the actual proof itself
Wow! This clarifies a good week or two from last year's lectures. I wish I had seen these videos when I was taking the course last year.
this is the best video ive seen on this topic, very well done
This makes so much more sense now, thank you!
The point of view in curvature is soooo great!
wow finally get the idea about this relationship between covariance matrix and hessian
so in otherowords the covariance matrix is hessian of maximum likellihood?
Wow, that makes things so much clearer. Thank you.
High curvature -> sharp -> concentrated -> low variance. Makes sense.
You just saved my semester (again) GGWP
This helps so much. very simple explanation
OK 3 months ago, I thought I understood this video. After I learned more statistic. Now I understand what is going on. I didn't quite understand the concept 3 months ago.
Kudos man! most intuitive explanation ever!
Ben, you are amazing!
Those tangents illustrate the convexity... Jensen!
Awesome awesome awesome video....Thankyou so much!
you said we add the negative sign, because the second derivative is negative after a certain value, and the negative sign is added to correct for that negative. what about when the second derivative is positive? doesn't the negative sign make the second derivative negative then? of what use will that be?
Isn't the variance of theta hat also dependent on n, the number of observations which constitute the likelihood function?
Wonderful video. Thank you very much!
thanks. Good explanation. I guess you saved me hours of searching.
Thank you so much. This explains so much.
Well explained man!!! Thanks a million 🙏
Excellent video, congratulations!
Thank you for this video. I have watched this video many times over the years. The simplicity, intuition, visuals, clarity, and ease, are nothing less than brilliant. It has always helped whenever things get fuzzy.
Just a small request or a question if you may: Calling vertical axis "likelihood of the data" makes it a bit confusing!
Instead, should it not be "likelihood of the parameter" that is L( theta; data). And this "likelihood of the parameter" then happens to be equivalent to f(data|theta)? So, y axis should not be called L(data|theta)?
Great video, as always. Helped me out a lot!
Thank u Ben, it was quite helpful
In wich playlist ı can find this topics in a ordered manner
Hi ben, thank you so much for your videos, i am studying quantitative ecology and do not have a strong mathematical background - your lessons really help! May I ask how the different values of theta are generated (along the x axis)? I assume the MLE expression stays constant and that the parameter estimates vary due to sample variation but in my case I only have one sample. I am a bit confused whether variance of the MLE is actually referring to variance in the parameter estimate due to sampling error. Secondly, in order to calculate the variance, must the 2nd derivative be evaluated for the value of theta which gives the MLE? I hope these questions make sense!
Awesome video!!
Hi Ben find your tutorials very easy to follow- thanks. What software are you using? Especially like the coloured pens on black background.
i dont know hat he is using but I sometimes use app.liveboard.online/ . It also allows you to chose different backgrounds for a board and different colors and to livestream your drawing from your tablet/smartphone to PC which i often use as it is better to draw by hand/pen then by mouse.
You can check his website for info.
wonderful video, thank you!
Many thanks, much appreciated!
Thank you very much!
Thanks, very intuitive.
[Subscribed]
You da best!
wonderful video
Excellent many thanks
great intuitive :)
Very good.
I want to know the meaning of penalized mle
Are you learning that for Machine Learning?
WOW!
ily
My student finance payment should be going to people like you, not these institutions.
شكرا و لكن و الله مفهمت 😂😂😂
نتمنى وضع ترجمة لاحقا
Kane Park
Jon Plaza
Lydia Stream
Thank you so much!