You know, I've never been able to convince myself that the quantization error is correctly taken into account by standard statistical error bars / power of study.