I need to calculate sample size for categorical data with three groups.
Proc Power and Proc GLMpower are using an alpha level of significance of 0.05 by default. I know that I can change it if I wish.
In our data we adjusted Fisher exact / Chi_square p-value by bonferroni method Now p-value is very small, like 0.0009, for example. Can I use this value as an alpha in Proc power? Are there any other SAS procedures that can calculate sample size using a Bonferroni p-value as a reference?
I think that you have some study re-design to do. A Bonferroni alpha level of 0.0009 indicates that you have 55 tests. A primary hypothesis (at least in a clinical trial) should be designed for just one test and maybe a few interim looks.