Validity and power considerations on hypothesis testing under minimization