Would be very interested in your take on the Gamma Distribution, especially in relation to its usage in Bayesian stats. It has strong connections to many of the ones you've covered recently, such as the Beta/Binomial and Dirichlet/Multinomial. There are some interesting algorithms for drawing from the Gamma distribution, such as the Chinese Restaurant process and the Stick-Breaking process. The Gamma Dist is a bit more difficult of a distribution, especially when it comes to writing your own code to calculate it, but it has so many important connections to other stats/dists that it shows up over and over again, especially when one wants to do things a little-bit more advanced than simple text-book problems.
the process comes from stochastic modeling. You can use gamma distribution as a way to generalize exponential distribution. Exponential distribution models the time between events which belong to a Poisson process. the gamma distribution models the time between Poisson process but u can have it model a particular trial. If you have a Poisson process with a rate of 2 events per hour, the time until the 3rd event follows a gamma distribution with shape parameter 3 and rate parameter 2. This means the waiting time until the 3rd event is the sum of three independent exponential random variables, each with a mean of 0.5 hours. With regards to bayesian statistics, we use the gamma distribution as an a priori to adjust parameters in a Poisson process. We can use gamma as a conjugate prior for Poisson likelihood. I suggest taking a course in stochastic models to learn more about this distribution, as that is where u can see a large amount of applications.
Hello @ritvikmath. I just want to say thank you so much for all the videos you are putting out here to help other people (like me). You are well appreciated. Keep up the good work👍👍👍
Is there some way to do exactly that but instead of the probability of exactly x1, x2 and x3 outcomes to have at least x1 and at least x2 outcomes? So: Given a trial with the possible outcomes x1, x2, x3 and x4 with the probabilities p1 = 0.1, p2 = 0.3, p3 = 0.15, p4 = 0.45 What is the probability of x1 >= 4 and x2 >= 3 within 15 trials?
Can we generally bypass the challenges of p-hacking and multiple testing corrections eg, Benjamini Hochberg, if we just use a good prob distribution to draw simulation results from, in a Bayesian solution format in lieu of p value and arbitrary significance level choices? I have seen nice things like this using a beta distribution. It seems to bypass so many old frequentist problems very neatly. Regards!
I had been planning to finally start my channel with exactly this topic, and here again you are, with a video that just cant be beaten. I just have no idea, whether I am too happy or too angry?
please start your channel! there's tons of videos on UA-cam about the same topic; different people learn differently so you'll definitely be adding to the community
Cheer~~~~an expression of more than two algebraic terms, especially the sum of several terms that contain different powers of the same variable(s)---- a polynomial.😊
Brilliant simplification of the 4 distributions . . . like a THEORY OF EVERYTHING of n and k . . .
Sooo glad you’re still active! Been watching your TSA and PCA playlists to prep me for grad school
You got this!
Would be very interested in your take on the Gamma Distribution, especially in relation to its usage in Bayesian stats. It has strong connections to many of the ones you've covered recently, such as the Beta/Binomial and Dirichlet/Multinomial. There are some interesting algorithms for drawing from the Gamma distribution, such as the Chinese Restaurant process and the Stick-Breaking process.
The Gamma Dist is a bit more difficult of a distribution, especially when it comes to writing your own code to calculate it, but it has so many important connections to other stats/dists that it shows up over and over again, especially when one wants to do things a little-bit more advanced than simple text-book problems.
the process comes from stochastic modeling. You can use gamma distribution as a way to generalize exponential distribution. Exponential distribution models the time between events which belong to a Poisson process. the gamma distribution models the time between Poisson process but u can have it model a particular trial. If you have a Poisson process with a rate of 2 events per hour, the time until the 3rd event follows a gamma distribution with shape parameter 3 and rate parameter 2. This means the waiting time until the 3rd event is the sum of three independent exponential random variables, each with a mean of 0.5 hours.
With regards to bayesian statistics, we use the gamma distribution as an a priori to adjust parameters in a Poisson process. We can use gamma as a conjugate prior for Poisson likelihood.
I suggest taking a course in stochastic models to learn more about this distribution, as that is where u can see a large amount of applications.
Love these distribution videos
Glad you like them!
Hello @ritvikmath. I just want to say thank you so much for all the videos you are putting out here to help other people (like me). You are well appreciated. Keep up the good work👍👍👍
I love this. I always absorb ideas more when it's framed through an absurd situation. I laughed when you introduced the cod fans LOL
Cleaned, and Detailed explnation Ritvik, thank you so much;
My pleasure 😊
I am doing MIT Micromasters in stats and your videos are godsend.
Going through these basic concepts are super helpful! I still go back to your video on kriging to understand the underlying science!
Glad it was helpful!
wondering if you can talk about copula in the near future especially the non Gaussian copula one. Btw, always love your video
Is there some way to do exactly that but instead of the probability of exactly x1, x2 and x3 outcomes to have at least x1 and at least x2 outcomes?
So:
Given a trial with the possible outcomes x1, x2, x3 and x4 with the probabilities p1 = 0.1, p2 = 0.3, p3 = 0.15, p4 = 0.45
What is the probability of x1 >= 4 and x2 >= 3 within 15 trials?
Amazing approach 👌 thanks a lot..
Can we generally bypass the challenges of p-hacking and multiple testing corrections eg, Benjamini Hochberg, if we just use a good prob distribution to draw simulation results from, in a Bayesian solution format in lieu of p value and arbitrary significance level choices? I have seen nice things like this using a beta distribution. It seems to bypass so many old frequentist problems very neatly. Regards!
I had been planning to finally start my channel with exactly this topic, and here again you are, with a video that just cant be beaten. I just have no idea, whether I am too happy or too angry?
please start your channel! there's tons of videos on UA-cam about the same topic; different people learn differently so you'll definitely be adding to the community
@@ritvikmath I agree. Sometimes you’re might get things from explanation in other videos things faster.
Amazing video
Thanks!
hi ritvik can you make a video about transformers and attention models please? its a subject i am having difficulties understanding
Cheer~~~~an expression of more than two algebraic terms, especially the sum of several terms that contain different powers of the same variable(s)---- a polynomial.😊
Thank you, very nice video
Glad you liked it!
Salmon team for the win! ;)
Haha 😂
Damn! Thank you
Of course 🙏
Thank you 👏😇
Of course!
Story is about a fish convention and he doesn't discuss the poisson distribution. Sounds a bit fishy.
🤨
This video is just one big tease 🤣😢