log in
register

[From my gist] AB testing

Qingqi@2020-07-07 #杂谈

post date: 2020-3-2

Also known as bucket testing or split-run testing, is a user experience research methodology.

A/B test is a randomized experiment with two variants, A and B. Are considered the simplest form of controlled experiment.

A and B have difference in a single variable, that might affect a user's behavior. The benefits of A/B testing are considered to be that it can be performed continuously on almost anything.

example

  • Landing-page optimization
  • Ad creative optimization

statistic tests

  • z-test
    • z = (p_A-p_B)/(\sqrt{SE_A^2+SE_B^2})
    • SE_A^2=p_A(1-p_A)/n_A
    • SE_B^2=p_B(1-p_B)/n_B
    • normal distribution
  • t-test (equal but unkonwn variances)
    • t = ((X1-X2)-(\mu1-\mu2))/(\sqrt{(s_p^2/n1+s_p^2/n2)})
    • df = n1+n2-2
    • s_p^2 is a weighted average of the sample variance
    • s_P^2=((n1-1)s1^2+(n2-1)s2^2)/(n1+n2-2)
    • t distribution
  • t-test (unequal and unkonwn variances)
    • t = ((X1-X2)-(\mu1-\mu2))/(\sqrt{(s_1^2/n1+s_2^2/n2)})
    • df = (s1^2/n1+s2^2/n2)^2/(s1^4/n1^3+s2^4/n2^3)
    • s1, s2 is the sample variance
    • t distribution
Comments

Log in to add your comment

Don't have an account? register here