A Guide to A/B Testing as a Data Scientist!
CODE: github.com/ajhalthor/bayesian...
REFERENCES
[1] Book on Bayesian Math: users.aalto.fi/~ave/BDA3.pdf
[2] Good blog to understand Beta Distribution: varianceexplained.org/statisti...
SPONSOR
Kite is a free AI-powered coding assistant that will help you code faster and smarter. The Kite plugin integrates with all the top editors and IDEs to give you smart completions and documentation while you’re typing. I've been using Kite. Love it! Learn more here:
www.kite.com/get-kite/?...
TIMESTAMPS
0:00 Introduction
0:47 Experiment Definition
3:00 Create Dataset for A/B Test
5:07 Hypothesis Testing
6:00 Issue of Interpretability
8:09 Bayesian Testing: Constructing Prior
10:07 Bayesian Testing: Constructing Posterior
13:55 Introducing a Continuous Metric: Prices
14:25 Understand Pricing Distributions
16:04 Add prices to dataset
17:05 Bayesian Math
20:37 Coded Math
22:47 Construct Priors, Posteriors
25:42 Interpret Results for Continuous Metric
Пікірлер: 11
As someone who is self-teaching himself ML/DL, I love your channel =]
@CodeEmporium
2 жыл бұрын
As a educator, I love this comment. Thanks for watching :)
For the first part of the video, just wanted to add a slightly more formal note to why beta distribution is a good choice for the prior distribution. It's because beta distribution is a 'conjugate prior' to the binomial distribution. The likelihood of observing a purchase conversion is represented by a binomial distribution here and the prior as a beta distribution. This gives us a posterior which is a distribution with the same functional form as the prior (i.e. beta distribution) with updated parameters to account for the information from the likelihood distribution. P.S. : Conjugate priors is a nifty concept in bayesian statistics.
Really Like the way you explain the topics in the easiest way.. ❤️.. Please keep it up.. 👍
Great work and great video! 😀
great content, hope to see more of this!!!
Hi, thanks for your videos I find them very informative. I have a query, in the frequentist approach, what is the reason for choosing for chi2 test?
Amazing job as usual Would be good to take the essential intuitions here Keep them in the same order Connect them at the visual level, 3B1B style you used to do at the beginning Then let me know about the number of hits on that video :)
I was quite confused with the H0: Control & Treatment are independent part, cause in my mind, if the treat and the control never shows together, we can think they are dependent , it must be a treat or non-treat, if you want to check whether get the treat & control not rand split , its base on other parameter like 'ever converted' , not the result converted
I wonder how you sampled from the posterior. It looks like you have sampled independently, am I right?
Please make a video custom voice text to speech