ABSTRACT
In the last chapter, we discussed the process of null hypothesis significance testing (NHST), which has been an integral part of all inferential statistics (e.g., t test, correlation, chi-square) in the biological, behavioral, and social sciences for most of the past century. However, concerns about its use and interpretation have recently increased for two reasons. First, the logic underlying NHST is difficult to understand, appears to be backward, and under certain conditions will always lead to rejection of the null hypothesis. Second, and perhaps more important, null hypothesis significance testing is often improperly used and interpreted in medical and behavioral research. Terms such as null hypothesis, significance level, statistically significant, and power, which are fundamental to NHST, are often used without an accurate understanding of their technical meanings. As a result, NHST appears to have been overused, and misinterpreted, so much so that many knowledgeable researchers advocated that NHST simply be banned. The purpose of this chapter is to (a) briefly present the logical underpinnings and proper use of NHST, (b) point out common problems accompanying NHST, and (c) make suggestions to increase recognition of misinterpretations of NHST in the published literature. Unfortunately, the topic is too complex to be dealt with completely in a single chapter, so we focus on the major points. However, for a more comprehensive treatment of NHST, we recommend the texts by Harlow, Mulaik, and Steiger (1997) and Kline (2004).
