very nice,
just, in my case, usually "transaction" data such as revenue, recharge, and the transaction itself that we want to validate follows a “not normal” distribution, maybe a gamma distribution? I don't know. they are right skewed. how do you handle the Tukey’s HSD under this assumption?