Thank you so much. Now it all makes sense. I had difficulty grasping the idea of MLE, but with your explanation I feel confident going back to the lectures and being able to follow them.
You all probably dont care at all but does any of you know of a way to log back into an instagram account?? I was stupid lost my password. I love any tips you can offer me
Thank you so much for your video, especially for warnings about model selection and AIC. Would you please explain more (or give me some documents or references) about "do not combine model selection with hypothesis testing. The p value significance will be inflated because you are implicitly testing multiple hypotheses with model selection"
isnt the (probability of x given theta) = (probability of theta given x)(probability x)/(probability theta) If this is the case, the "likelihood function" as you defined, is it really equal to the probability of x given theta ? If so, why, since it is missing those two terms extra terms?
That is true in a Bayesian setting, where the parameters in a given model are treated as being random. The point of maximum likelihood estimation, on the other hand, and Frequentist inference is we treat parameters as being static, such that we can estimate them. I wish the author had written in the more common notation for density for MLE, with f(x ; theta) instead of f(x given theta), it's confusing when this isn't known. Btw you are correct, so Bayes theorem works for densities too! p(theta) is the density of the parameter, p(theta given x) is the parameter conditioned on the data (the thing we want!) and p(x) the normalizing constant. Bayesian inference is basically the science of picking a prior based on objective/subjective mathematical means.
AIC - the lower the better, LL - the higher the better, but both measure the same concept, so using both is a redundancy, one will suffice (as one will always go down when the other goes up judging by the formula). Did i get it right?
Best MLE video on TH-cam! Thank you :)
Agree! Thank you also from my side!
Great explanation. Clear, structured and explained in simple understandable terms. Thanks for taking the time to put this together.
Most clear explanation i have seen on TH-cam thus far
Thank you so much. Now it all makes sense. I had difficulty grasping the idea of MLE, but with your explanation I feel confident going back to the lectures and being able to follow them.
Matthew you are outstanding as a teacher. Thank you for the many insights and teaching.
-Steve G.
You all probably dont care at all but does any of you know of a way to log back into an instagram account??
I was stupid lost my password. I love any tips you can offer me
Matthew you are awesome.
I wish you did a video on Bayesian too. Bayesian, MCMC one please??
Thank you so much for your video, especially for warnings about model selection and AIC. Would you please explain more (or give me some documents or references) about "do not combine model selection with hypothesis testing. The p value significance will be inflated because you are implicitly testing multiple hypotheses with model selection"
Fantastic! Thank you so much for this super clear exposition.
Really nice presentation
بارك الله فيكم وجزاكم الله خير الجزاء
Thank you so much!! Such clear explanations!!
Nicely explained, thanks!
Yeaahh thats the clear explanation
Thank you for taking the time to make this video.
Great video man! Helped me a lot, all the best :D
Seriously good!
Really it is very interesting!! Thank You!!
Thank you very much for your introduction!
Thank you very much!
Amazing, I finally understood MLE
you're amazing sir
The last slide is gold
Very helpful
isnt the (probability of x given theta) = (probability of theta given x)(probability x)/(probability theta)
If this is the case, the "likelihood function" as you defined, is it really equal to the probability of x given theta ?
If so, why, since it is missing those two terms extra terms?
Fahraynk I have the same question, please tell me when you found an answer.
That is true in a Bayesian setting, where the parameters in a given model are treated as being random.
The point of maximum likelihood estimation, on the other hand, and Frequentist inference is we treat parameters as being static, such that we can estimate them. I wish the author had written in the more common notation for density for MLE, with f(x ; theta) instead of f(x given theta), it's confusing when this isn't known.
Btw you are correct, so Bayes theorem works for densities too! p(theta) is the density of the parameter, p(theta given x) is the parameter conditioned on the data (the thing we want!) and p(x) the normalizing constant. Bayesian inference is basically the science of picking a prior based on objective/subjective mathematical means.
AIC - the lower the better, LL - the higher the better, but both measure the same concept, so using both is a redundancy, one will suffice (as one will always go down when the other goes up judging by the formula). Did i get it right?
How does restricted maximum likelihood estimation change the description here?
cool! Good job!
this is great!
AWESOME!
Thank you!
thanks
It is interesting to me why they just do not divide AIC eqn. by 2.
nice explanation...but i ended up cleaning my laptop screen.. after 4.49
Thankyou sir. Its very helpful. Can you please show the mathematical workout of this in figures?
Is likelihood same as probability?