Most of times we forget to check multivariate normality Mardia's kurtosis. If the data cannot meet the normality requirement, we can run bootstrap to get bootstrap p which is generally bigger than p with MAL. Always put theory in mind and dont become a pure dataminers.