I have another question. Group sizes are unequal in my sample because of some practical reasons. What would be the risk if the assumption of homogeneity of variances is violated in this case?