What is the minimum length of a time-series for considering the results of the Mann-Kendall trend test "coherent"?

  • $\begingroup$ What do you mean by "coherent"? $\endgroup$ – jbowman Dec 12 '11 at 23:10
  • $\begingroup$ After I posted this, I found that in order to consider Z following a normal distribution the length of the time series should be over 10 $\endgroup$ – papafe Dec 13 '11 at 14:14
  • $\begingroup$ 1. How does the quality of that approximation relate to "coherence" (whatever coherence mean)? 2. The R package doesn't use the normal approximation for Kendall's tau until n exceeds 50. The approximation is a bit rough at n=10, Personally I would avoid using it until n was a bit nearer 20. $\endgroup$ – Glen_b Jun 5 '16 at 9:12

The Mann-Kendall function in R requires at least 3 data points. Look here: Mann-Kendall test in R using 'Kendall' package: exact test or approximation?

| cite | improve this answer | |
  • $\begingroup$ I am not sure this answers the question asked, since that concerned whether the results could be defined as "coherent", not just whether the calculation could be performed. But sadly the intended meaning of "coherent" was not well-explained. $\endgroup$ – Silverfish Feb 20 '16 at 21:01

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.