When you violate an assumption underlying a statistical test, it doesn't mean that the data are bad. It's an effect in your data.