Dear Allstat,
I have 4 proportions p(A),p(B),p(C) and p(D) that I would like to compare, and was wondering the best way to go about this. I can perform a Chi^2 test but that just tells me that there is a difference (or not) amongst the 4 proportions for a given alpha. I could also look at the Pearson residuals for further info.
But I am really interested in identifying which of the 4 are different so a bunch of pairwise comparisons. My thought is to use the standard Normal approx to compare proportions and adjust my alpha for the multiple comparisons (Bonferroni correction). Does this approach make sense or are their better methods?
This approach seems analogous in some ways to an ANOVA test for comparing multiple means followed by multiple comparisons based on the LSD (Least Significant Difference)?
Any thoughts, help, references most appreciated
Best Regards,
Mary.
Some background. A, B, C and D are samples from 4 population segments and p(A), p(B), p(C) and p(D) are the proportions of "yes" responses to an on line survey question.
|