What If Null Hypothesis Significance Tests Were Banned? An Interactive Discussion of Implications for General Linear Modeling

• In Event: Multiple Linear Regression: The General Linear Model SIG Business Meeting

Abstract

There has been longstanding controversy around the use of Null Hypothesis Significance Testing for decades (e.g., Cohen, 1994; Harlow, Mulaik, & Steiger, 1997). There is arguably renewed vigor around this controversy with the ever expanding access to large datasets (i.e., big data) and the recent banning of the procedure in the Journal of Basic and Applied Psychology (Trafinow & Marks, 2015). Related dialogue should account for recognition that even in the context of census-level data, there is often some desire for making inferences about other contexts and settings, and some applications of big data entail conceptualization of potentially infinite or otherwise unmeasurable populations (i.e., a true census is not possible). There should also be discussion of applications dealing with non-nil nulls, replicability, exact p values, and so on. These applications should be considered in the context of the GLM; therefore, the speaker will lead an interactive discussion of GLM applications while maintaining focus on research purpose and questions.

Authors

• Isadore Newman, Florida International University

• John H. Hitchcock, Indiana University - Bloomington

• David O. Newman, Florida Atlantic University

• David A. Walker, Northern Illinois University

• Thomas J. Smith, Northern Illinois University

• William Holmes Finch, Ball State University

• Gordon P. Brooks, Ohio University - Athens