What is an acceptable probability of sampling error that may have skewed the results in most scientific studies?