3

I have a question about wilcoxon signed rank test. I have 2 different speed profiles (speed starts from time 0 to 100 sec) of the same participant that was given 2 different treatment. I would like to ask if I could use wilcoxon signed rank test to compare if both speed profiles are different from each other. I am a little bit confused how to fit the null hypothesis of this test (median difference) to my case. I also think that once I use wilcoxon signed rank test to my time-series data, I may lose the sence of data sequence from my speed profile.

Any help from anyone will be appreciated.

Thank you, Ayla

user46353
  • 31
  • 2
  • 1
    Wilcoxon implies pairs of observations are independent, which is probably not so with your time series data. So the answer is no. – ttnphns May 29 '14 at 06:26
  • @ttnphns: I read this comment on a similar question http://stats.stackexchange.com/questions/78788/can-i-use-wilcoxon-signed-rank-test-to-test-pairs-of-data-in-a-time-series. Glen_b answered that this test can be used if I want to test for distribution. Can you please elaborate more? Thanks – user46353 May 29 '14 at 14:14
  • @ttnphns: do you mind if i ask what the alternative test that you think might fit for my analysis? thanks – user46353 May 29 '14 at 14:19
  • @user46353 also the null hypothesis of a rank sum test is that that there is no stochastic dominance. The test for median difference is a special case that only holds when the shape of the distributions in both groups is identical, differing only by a shift in centrality. – Alexis May 29 '14 at 15:04
  • @user46353, Sorry, I'm not an expert in time series analysis. Try to google search "comparing two time series". I expect that the task is not very easy. You can also ask a new question on this site. Also, be aware that I didn't quite understand from your description whether your data are time series or not. I only said that Wilcoxon requires that paired observations are collected independently. – ttnphns May 29 '14 at 16:56
  • @user46353 That's a clear misrepresentation of what I say at the link, the first paragraph there is completely in keeping with the first comment ttnphns made. I make it quite clear that the rest of the comments are conditional on there being independence; as ttnphns says, it's probably not so with your time series data. – Glen_b Apr 05 '17 at 05:46

0 Answers0