I am running a chi square test, followed by a marascuilo test to compare a binary resonse with categorical variables as the independent variables. The dilema is that my total sample size is 1,000,000 so I am getting chi square values in the thousands, with no major differences in the actual proportions. For example, there were four subgroups with % successes of 53, 42, 55 and 47 and I got a chi square statistic of 25,000. Is there any way to control or account for such an enormous sample size. With this sample size, I'll be finding sig diffs that are only 1-2% diff, likely even less. Any suggestions.