Statistics for Verifying Explanations for Neural Networks