I have a 2 x 3 repeated measures design with significant interaction, and I used LSD as a post hoc for the interaction but it does not adjustment the alpha. Which correction would be the most conservative one for this design or any other post hoc that can be run?