Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Next revision
Previous revision
data_analysis:ab_test [2021/02/07 14:56]
prgram created
data_analysis:ab_test [2021/09/30 10:07] (current)
prgram [A/B test]
Line 1: Line 1:
-====== ​ab test ======+====== ​A/B test ====== 
 + 
 +AB 테스트 추천 도서 : [[https://​www.amazon.com/​Trustworthy-Online-Controlled-Experiments-Practical/​dp/​1108724264|Trustworthy Online Controlled Experiments:​ A Practical Guide to A/B Testing]] [[https://​coupa.ng/​b7Omnt|쿠팡]] 
 +통계적 지식 뿐 아니라 실제 적용에서 고민해야할 문제들에 대한 상세한 설명이 있다. 
 + 
 + 
 +[[data_analysis:​convergence_testing]]
  
 https://​reflectivedata.com/​comprehensive-guide-to-statistics-in-a-b-testing/​ https://​reflectivedata.com/​comprehensive-guide-to-statistics-in-a-b-testing/​
-https://​cdn2.hubspot.net/​hubfs/​310840/​VWO_SmartStats_technical_whitepaper.pdf 
 https://​www.slideshare.net/​cojette/​ab-150118831 https://​www.slideshare.net/​cojette/​ab-150118831
 https://​onlinemix.tistory.com/​entry/​significant-result-from-ab-testing https://​onlinemix.tistory.com/​entry/​significant-result-from-ab-testing
-https://​is.muni.cz/​th/​wt0tu/​Humaj-thesis.pdf 
 https://​cxl.com/​blog/​ab-testing-guide/​ https://​cxl.com/​blog/​ab-testing-guide/​
 https://​hbr.org/​2017/​06/​a-refresher-on-ab-testing https://​hbr.org/​2017/​06/​a-refresher-on-ab-testing
  
 +Bayesian
 +https://​is.muni.cz/​th/​wt0tu/​Humaj-thesis.pdf
 +https://​brunch.co.kr/​@gimmesilver/​15
 +[[https://​cdn2.hubspot.net/​hubfs/​310840/​VWO_SmartStats_technical_whitepaper.pdf|VWO_SmartStats_technical_whitepaper]]
 +
 +
 +[[https://​www.researchgate.net/​publication/​287927836_Sample_Size_Calculation_for_Two_Independent_Groups_A_Useful_Rule_of_Thumb|Sample size calculation for two independent Groups]]
 +
 +
 +https://​www.evanmiller.org/​ab-testing/​
 +
 +
 +MAB
 +http://​sanghyukchun.github.io/​96/​
 +https://​kw94.tistory.com/​49
 +https://​sumniya.tistory.com/​9
 +
 +
 +===== 실행순서 ====
 +  * 가설 : 인터뷰, idea, 기존자료 분석
 +  * 실험설계 : control/​treatment,​ 지표, bayesian/​frequentist
 +    * Random unit : user/​session/​page
 +    * Target unit : 전체 / segment
 +    * Size : sample, traffic 분배
 +    * How long : 학습효과
 +  * 실행 : 모니터링. 오류, 다른 feature들이 떨어지는지,​ 예상치 못하게 효율이 떨어지는지 (cf. 가드레일 metric)
 +  * 분석 : traffic 분배 정합성, 샘플수, 다른 segments의 특성, 다른 지표, Bot은 없는지
 +  * 결정 : 통계적vs실질적 유의도, A/A test->​검증
  
 {{tag>​data_analysis 실험 experiment ABtest}} {{tag>​data_analysis 실험 experiment ABtest}}