Hi everyone,
I have a large set of individual likert items (they don't consitute scales) and two independent groups. Now I want to exploratory test whether the groups differ in their responses on the likert items. I've considered using a Wilcoxon rank-sum test. But unfortunately the distributions often differ between the groups, so I'm not shure about whether this gives the adequate result. As far as I understand Wilcoxon tests for differences in medians if the variability is similar across groups. How would I test for differences in central tendencies in this case?
'
Pls take a look at mann-whitney and kruskal-wallis tests. If you have any problem in R contact me on https://www.fiverr.com/s/1o3Y4z
Alamir Louro , isn't the Mann-Whitney test equivalent to the Wilcoxon rank sum test mentioned in the question ?
Stefan Becker , you're correct that for the Wilcoxon rank sum test to be a test of the medians, the distributions (of the underlying population) distributions need to have the same shape and spread. But without this condition, the Wilcoxon rank sum test is a test of stochastic equivalence. That is, if one group tends to have higher values than the other group. This hypothesis is often of interest.
Dear Sal, your are right, the Wilcoxon test and the Mann-Whitney U test are equivalent, but the Wilcoxon test function help in R say that the literature is not unanimous about the definitions of the Wilcoxon rank sum and Mann-Whitney tests. And R implement both using the same wilcox.test (see the function help in R).
I cited to use mann whitney and kruskal wallis because I dont know how the problem is, but one of both probably it is the solution. depends on how many independent variables you have.
OBS: If there is many dependent variables, manova.rm could help.
OBS: invariance tests could be a good step too.
BR
Thank you both Sal and Alamir,
in fact I did the Wilcoxon rank-sum test in R with wilcox.exact (and SPSS to compare).
I'm performing multiple group comparisons. Partly samplesizes are below 20.
Therefore I've set exact = TRUE in these cases. Unfortunately the p-values then are different from the exact ones SPSS gave. Do you have any idea on this?
By the way, the mean ranks indicate which group has scored lower/higher on the items, right?
It is complex to compare tools' precision. Yes, the test results present if there is difference and what is the mean of which group.
- Badges
- Science topic
- Similar topics
- Filing