I just learned that most tests of associations on 2x2 tables have assumptions about row and column margins being fixed or not.

Is there any similar assumption when it comes to chi2 tests? Can I do it if my row-margins are fixed? (Example: Experiment, where I assign participants to predefined treatment groups)

  • $\begingroup$ What do you mean by row and column margins being fixed? $\endgroup$ Aug 6, 2019 at 11:52
  • $\begingroup$ I mean fixed marginal totals. For example, in an experiment, if I assigned 30 people to treatment A and 30 to treatment B, I would have fixed marginal totals in the rows (30/30) but not fixed marginal totals in the columns. $\endgroup$
    – Poza
    Aug 6, 2019 at 12:20
  • 1
    $\begingroup$ At stats.stackexchange.com/a/17148/919 I provide a general account of assumptions needed in all chi-squared tests. $\endgroup$
    – whuber
    Aug 6, 2019 at 12:33
  • $\begingroup$ One thing people often appear confused about is the difference between believing the margins to be fixed and conditioning on the (almost ancillary) margins. One needn't think margins are fixed to condition on them. $\endgroup$
    – Glen_b
    Aug 7, 2019 at 4:02


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy