I have 2 independent variables (gender of participant and gender of victim) both have two levels (male and female).

I have 4 dependent variables and each dependent variable had 3 5-point liker scale questions relating to it. Am I right in assuming I find the median score for each participant on all 3 questions for each dependent variable so I have 1 overall score for each dependent variable?

I am also using age and education as possible covariates.

I've started running the assumptions for two-way mancova and straight away my p values are statistically significant, there are 3 extreme outliers and a normal distribution cannot be assumed.

What can I do?!