I wonder if there is a possibility of doing power analysis for post-hoc test for GAM? I am specifically interested in a sample size necessary to achieve a desired power.
