This paper addresses the appropriate unit of analysis in field research. We first discuss the issues related to this topic: (a) unit of measurement versus unit of analysis, (b) treatments and random assignment, (c) independence of observations, (d) moderating and control variables, and (e) correlational versus experimental research. We then present a model for determining the correct unit of analysis. In many instances, researchers should use class means or subgroup means, and this has implications for research design. In the third section, we discuss the related issues of (a) the burden of proof, (b) asking the right questions and getting the right answers, and (c) completing statistical analyses. How data are analyzed can affect the results, and researchers should consider these issues when planning their research.
Stephen Silverman is with the Department of Kinesiology at the University of Illinois at Urbana-Champaign, Louise Freer Hall, 906 South Goodwin Ave., Urbana, IL61801. Melinda Solmon is with the Department of Kinesiology at the University of Maryland at College Park, College Park, MD 20742.