118 – SPEED: Teaching Statistics: Strategies and Applications
Type S Error Control in Hypothesis Testing
Andrew Neath
Southern Illinois University Edwardsville
Hypothesis testing is commonly introduced with the problem of testing for a difference between treatment groups. A valid criticism of hypothesis testing problems in this setting is that all null hypotheses are wrong on the scientific grounds that treatments always have some difference in their effects. But there is a very simple answer to this criticism. If we instead consider the problem of testing for the direction of the treatment difference, a hypothesis test can be developed by controlling for the probability of a type S (or sign) error. A type S error occurs when the test claims the treatment difference is in the positive direction when the true direction is negative; or when the test claims the treatment difference is in the negative direction when the true direction is positive. Presenting a hypothesis testing problem in this fashion allows for a more meaningful introduction to the best uses of statistics in advancing scientific knowledge.