Comparing groups that are not independent

[BettyH]

I work with survey data and am interested in comparing the responses of different racial/ethnic groups. So for example, do the responses of White, Asian, Black, and Latinx differ significantly on question #1 in my survey. Race/ethnicity, however, is a "select all that apply" question. So, the same individual can appear in multiple groups. This doesn't seem to be an independent samples test but it is not a paired test either. Any advice on the appropriate nonparametric test(s) to run would be appreciated.

 

[hlsmith]

As long as a person isn't contributing data to more than one group in your analyses you should be fine. So you will have to create a rule to assign people with more than one listing to a single group. Then you should think about how this misclassification "rule" may impact your results.

 

[fed2]

isn't this called 'upcoding' sometimes?