• Random samples eliminate bias – but not variability.
• If the variation when we take repeat samples from the same populations is too great, we can’t trust the results of any one sample.
• But, taking many repeat samples, the variation from sample to sample will follow a predictable pattern.