How should a projected hypothesis test be modified if you're particularly concerned about

the type I error?
the type II error?