# Measure of Association between two Ordinal Variables

I have two ordinal variables. One is independent and the other one is dependent variable. I would like to test if there is monotonic relationship between these two or not. I am using Somers' d measure in SPSS.
Now to use Somers' d measure, does it matter if more than 20% of the cells have expected frequencies below five?
I think this assumption only applies to the Chi-square test for the nominal by ordinal variables or two nominal variables. I am not 100% sure!

• Does each specimen get a rank or get categorized into an ordinal scale? If the first, then @Lan's suggestion of the Spearman makes sense. If the second, then something like Somer's D is more appropriate.
– Ashe
Mar 22, 2016 at 18:58