1
$\begingroup$

I collected data on one sample, the DV could be separated into two groups (success yes vs. no) and then I have several IVs with interval scale. I just don't know if to use Wilcoxon or Man-Whitney test. Also I don't know if it's necessary to use Bonferroni correction or if that is just important for parametric tests.

$\endgroup$
  • $\begingroup$ What do you want to test? What is your hypothesis? $\endgroup$ – Baumann Feb 18 '14 at 14:42
  • $\begingroup$ @Baumann my original plan was to run a logistic regression and identify those variables that are best to predict the criteria. that didn't work for several reasons - my plan now is to do this by using correlation and/or parametric-tests. does that make sense? $\endgroup$ – Jennifer Feb 18 '14 at 19:27
1
$\begingroup$

From https://www.beds.ac.uk/howtoapply/departments/psychology/labs/spss/Mann-Whitney_U_and_Wilcoxon:

The difference between the Mann-Whitney U and the Wilcoxon tests relates to the design of the >experiment. If your experiment has a repeated measures or matched participants design then >the Wilcoxon test is used to analyse your data. If your experiment has an independent >measures design then the Mann-Whitney U test is used to analyse your data.

$\endgroup$
1
$\begingroup$

IMHO @Giorgio's anwser is not correct since there are two Wilcoxon tests (namely, Wilcoxon Rank-Sum and Wilcoxon Signed Rank).

It might be useful to remember that Mann-Whitney test is also called Wilcoxon rank-sum test. And therefore, there is no difference in applying Mann-Whitney or Wilcoxon rank-sum tests. You should get the same result.

As I already wrote here, people do sometimes make a mistake in replacing Mann-Whitney with the Wilcoxon signed-rank test. The difference comes from the assumptions. In the Mann-Whitney test you are interested in the difference between two independent populations (null hypothesis: the same, alternative: there is a difference) while in the Wilcoxon signed-rank test you are interested in testing the same hypothesis but with paired/matched samples.

For example, the Wilcoxon signed-rank test would be used if you had replicates (repeated) measurements between different time points/plates/... since it is the same sample but measured in different time/on different plates.

As for the Bonferroni correction, the answer to this question highly depends on your setting. If you are going to torture your data until it confesses I would advise that you use a correction technique (like Bonferroni/FDR/Holm-Bonferroni/...) to protect yourself from false judgement. Regardless of the choice between parametric and non-parametric tests.

$\endgroup$

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.