I have a one way unbalanced anova say with a levels.
My task is to find out if we have significantly different COUNTS at the various a levels.
I tried this : I compute the sum of the number of cases at each level a say $n_i$
How do I compute if the $n_i's$ are significantly different ? I do not have a measure of the internal variation in the $n_i's$
What am I missing?