Randomization Units in A/B Testing: Easy Explanation for Data Science Interviews

[@roha3111]

If videos are in correct order of learning then the playlist is more beneficial for data enthusiast. I hope Emma will do it.

[@VandanaVandana-y5e]

Hey Emma, Quick question: What exactly is the coarseness? Could you give a brief explanation .

[@pingdingdongpong]

A user can see 2 different versions of a page, therefore it is coarser. A page can load 2 different sizes of buttons, therefore it is coarser.

[@MaraLearns]

Hi Emma! Thank you for such a helpful video! But what is the best option of randomization unit for the experiment where we compare two ad campaigns (for example, sephora from the previous video)? If our users aren't authorized, should we choose cookie within specific amount of time? or pageviews? given that the driver metric is click thought rate

[@SumedhaNagpalxyz]

Hi Emma, Very informative video! Would you have suggestions on how to select randomization units for B2B products like data bricks and slack, which have a many to 1 relationship between users and accounts?

[@keshavgupta308]

Hii Mam . I love the way you teach.. Your way of teaching is even better than my college teachers... Things that I can't able to understand in my class I understand here in your videos....... Once again Thanks A Lot Mam .👌🙌

[@emma_ding]

Happy it put a smile on your face Keshav!

[@keshavgupta308]

@@emma_ding mam i want the personal tips regarding this Please try to contact of it is possible I am watching your videos from very old time. 🙏🏻🙏🏻

[@ds_yat5918]

Thank you so much Emma for such a great video!! For future video contents, do you mind talking about some concepts such as diff in diff model, experiments in two-side marketplace, and so on? Thank you so much!!

[@emma_ding]

Great suggestion! Thank you so much!

[@shakeeljessa9619]

Another great video Emma

[@emma_ding]

Thank you! 😀

[@michaelq4261]

Thank you. Lots learned. As always.

[@nothingspecific314]

Are page level metrics measured at the user level ex: CTR(user) = (num clicks by user)/(num impressions) or for the population (ex: CTR = sum(clicks)/sum(impressions))?

[@hEmZoRz]

It's possible to do it both ways. With the first one you'd then take an average across all users to get the population-level metric. Though, only the first approach would fulfil the assumptions of a t-test regarding observations' independence, so that's a better choise.

[@vishalranjan7261]

I am fresh learner in data science field , please provide guidance or road map to become data scientist , please help 🙏🙏

[@emma_ding]

Hi Vishal, thanks for your comment! 💛 Feel free to check out our website at www.datainterviewpro.com/ for information and resources on how to land your dream job as a Data Scientist.

[@pinyichang4680]

good