1

if you were comparing two randomly assigned groups of people for similarity at the start of a study, which statistical test would typically be best for:

  1. Marital status
  2. Height
  3. Likert item statement - I like chocolate
gung - Reinstate Monica
  • 145,122
s. stamos
  • 11
  • 2

1 Answers1

0

Marital status (assuming it is binary): usual choices are chi-squared or g-test

Height: t-test, assuming it is sufficiently Normal/bell-curved. If not Mann-Whitney

Likert: depends on the distribution. If you can bin it into 2 groups (e.g. "agree" and "strongly agree" vs everything else) then go with chi-squared, if you have many points on the scale and it's "sufficiently bell-shaped", you can get away with t-test. Otherwise you have to use something non-parametric, like Mann-Whitney.

pavel
  • 101
  • 1
  • 1
    None of those tests will demonstrate similarity ("similar" would be small effect size -- at least for certain kinds of effect size -- not large p-value). – Glen_b Jan 27 '16 at 16:24