Hi everyone,
My data consists of multiple ordinal items (5 point likert items) as dependent variables and several nominal grouping variables with different numbers of categories.
For each of combination of the ordinal item and grouping factor I've run a separate KW test and Dunn's tests with Bonferroni correction for ordinal items that produces significant results in KW.
My concern is that the distribution shapes of the responses on the dependent variables differ quite much between the levels of the grouping variables. Does this affect the validity of the results?
It doesn't matter as long as you use non-parametric tests, since they take different distribution shapes into account. Most researchers just test for normality, not any other distributions anyway. So perhaps don't even mention those distributions if you don't need to. But I can see you like to be conscientious :)
Hey Andrew, thanks for your kind advise. I suppose you're right about me being a bit overconscientious :)
- Badges
- Science topic
- Similar topics
- Filing