https://www.growthbook.io/ logo
#ask-questions
Title
# ask-questions
b

best-application-57321

10/27/2023, 9:01 AM
Hello,I noticed that you are using T-tests for the conversion rate A/B testing with frequentist statistical methods. Why not use a Z-test?
1
r

rhythmic-agent-34208

10/27/2023, 9:01 AM
@blue-sundown-32899 has marked the issue as closed. Reopen Issue button
b

billions-xylophone-11752

10/27/2023, 11:02 AM
Hey, there - I'll pose this question to our data science team and get back to you!
b

best-application-57321

10/27/2023, 6:52 PM
Thanks
r

rhythmic-agent-34208

10/30/2023, 7:27 PM
Hi there, I was able to chat with our Lead Data Scientist and he passed along the following: One should usually never use a z-test if you can use a t-test. The difference between which test to choose has to do with whether you know the population variance or whether you're estimating it. In practice, we're always estimating it and a t-test is consequently more appropriate. The difference between the 2 tests disappears for basically any meaningful, online A/B test.
I hope this answer your question satisfactorily, but if you have any follow-up questions just let me know here in this thread. Thanks!
b

best-application-57321

10/31/2023, 9:31 AM
thanks
3 Views