Seacucumbers were counted at the same 63 survey sites in three successive years (3 x 63 = 189 surveys in total). The distribution of counts per site are non-normal for each year. (Some sites have loads of seacucumber while most sites have few or none...they seem to have very patchy distributions). I would like to tell if there is a statistically significant difference between years for each of the 63 sites. I would also like to pool the sites into groups and compare if there is a significant difference between years for groups.

I think I need a non-parametric test for repeated measures. The only thing close I can find is a Kruskal-Wallis one-way ANOVA, but this test requires that samples are independent, which they aren’t because the same sites are re-sampled.

I would be very grateful for suggestions on the best statistical test for this situation.

  • 1
    $\begingroup$ When you have three dependent observations, they're not referred to as paired (that's two). In the situation you have, you might call them repeated measures. (You also have count data; you might want to consider suitable models for count data.) $\endgroup$
    – Glen_b
    Mar 29, 2015 at 9:45
  • $\begingroup$ @Glen_b Sure, I have edited the question to reflect that they are in fact 'repeated measures' as you point out. Do you have any advice on what type of models may be suitable for such count data? Thanks. $\endgroup$
    – Elizabeth
    Mar 29, 2015 at 10:00
  • 1
    $\begingroup$ Why do you think you need the test to be nonparametric? $\endgroup$
    – Glen_b
    Mar 29, 2015 at 10:13
  • $\begingroup$ Do you want to compare the three counts separately for each site (i.e. using 63 statistical tests)? $\endgroup$
    – Michael M
    Mar 29, 2015 at 12:57
  • $\begingroup$ @Glen_b The distribution of counts by site by year have non-normal distributions, hence a nonparametric statistic needs to be used. $\endgroup$
    – Elizabeth
    Mar 29, 2015 at 13:57


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.