Want to ace your data science interview? 🚀 85% of successful candidates practice with real interview questions. Explore the ones companies actually ask: www.interviewquery.com/questions?
F-test is not used for individual coefficients in a regression model. T-test is what is used for individual coefficients, f-test for the entire model. Hanif is incorrect here 20:00, somebody should have fact checked this.
For ALL the extracurricular activities, feature importance determination and what not, makes one wonder why setting the original constraint of not being able to use A/B testing and how realistic that constraint is?
@@justinc5043it is realistic in many industries. I work in health care and it’s infeasible to run experiments a lot of the time (ethics, costs, etc.). Quasi experiments and observational studies. Propensity Score Matching all day.
For measure the success, I would like to go with CX metrics like CSAT, NPS, CES, OSAT, etc. Moreover, why not social media monitoring tools like Brandwatch to understand the customers sentiment and measure success. She is talking more in data science concepts like p-value, algorithms, statistics, etc. Questions are not that great too.
The method in measuring impact based on matching doesn’t sound right to me. The so called control group that do not use stories is not a good control since they may see it and decide not to use it on purpose. The users that are not using stories will be smaller and smaller over time, making the group even more biased.
Can use a group of users where Stories isn't made available to them so there's ideally no awareness hence no degradation/cross contamination/bias but I cannot fathom what is a good rationale for purposely NOT going the A/B route (hence a fundamental flaw in the underlying premise of the question).
Want to ace your data science interview? 🚀
85% of successful candidates practice with real interview questions.
Explore the ones companies actually ask: www.interviewquery.com/questions?
F-test is not used for individual coefficients in a regression model. T-test is what is used for individual coefficients, f-test for the entire model. Hanif is incorrect here 20:00, somebody should have fact checked this.
For ALL the extracurricular activities, feature importance determination and what not, makes one wonder why setting the original constraint of not being able to use A/B testing and how realistic that constraint is?
@@justinc5043it is realistic in many industries. I work in health care and it’s infeasible to run experiments a lot of the time (ethics, costs, etc.). Quasi experiments and observational studies. Propensity Score Matching all day.
Its a great interacting video , can aspirant for Da who wants to get an idea of real work or hands on exp should go thru .
really interesting points made by Priya! thanks, good learning.
Thank you! We hope this helped!
For measure the success, I would like to go with CX metrics like CSAT, NPS, CES, OSAT, etc. Moreover, why not social media monitoring tools like Brandwatch to understand the customers sentiment and measure success. She is talking more in data science concepts like p-value, algorithms, statistics, etc. Questions are not that great too.
Because those metrics are really broad and don't let the user behaviour drive your understanding of the situation.
The method in measuring impact based on matching doesn’t sound right to me. The so called control group that do not use stories is not a good control since they may see it and decide not to use it on purpose. The users that are not using stories will be smaller and smaller over time, making the group even more biased.
Can use a group of users where Stories isn't made available to them so there's ideally no awareness hence no degradation/cross contamination/bias but I cannot fathom what is a good rationale for purposely NOT going the A/B route (hence a fundamental flaw in the underlying premise of the question).