Dealing with Inherent Statistical Error

Abstract

When performing an inferential statistical test, one can never be absolutely sure that the results are correct. Through careful study design and with good scientific technique, systematic error can be controlled, but random error will always be present. This chapter focuses on hypothesis testing and types of random error associated with incorrect decisions: (1) rejecting a true hypothesis under test or (2) failing to reject a false hypothesis under test. The goal is to reject the hypothesis under test with at least 95% confidence in the decision or less than a 5% chance of being wrong. Another goal is to reject the hypothesis under test when that hypothesis is false, which is a called statistical power. Minitab applications to determine sample size and power are discussed.