Jump to content

eros

S4GRU Member
 View Profile  See their activity

  • Posts

    1

  • Joined

  • Last visited

 Content Type 

Everything posted by eros

  1. eros
    It's a matter of a biased sample, as dallas suggested. You only need 191 people to estimate a moderate effect size in a 4x2 table (goes up to 1700 with a small effect size). The population size is irrelevant (see: central limit theorem) given an unbiased sample. A proper analysis would have involved taking current subscriber counts and assigning weights based on the users who downloaded Slice (prior to entering iPhone tracking information), then running the same table with the new weights.
×
×
  • Create New...