I have population data on four groups. Every member of each
group is given a grade A, B, C or D. The grade is ordinal in
nature A>B>C>D. The data then forms a 4x4 contingency
table.
Some of the cells in the table have frequencies of zero. To
obtain frequencies >5 would require combining across groups
and grade.
I wish to test if the grade is independent of the group. Is the chi-
squared test the most appropriate? A colleague has suggested
using the Kruskal-Wallis test because of the ordinal property of
the data. But from what I can gather this assumes the data
forms independent samples, does this matter?
Thanks
James
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|