ABSTRACT
Over the years, there have been numerous articles in the organizational and social sciences discussing various types of chopped data (i.e., continuous data that are partitioned into far fewer categories for data analytic purposes). These have included discussions on the practice of dichotomizing (or polytomizing) independent and dependent variables (Cohen, 1983; MacCallum, Zhang, Preacher, & Rucker, 2002), moderator variables (Bissonnette, Ickes, Bernstein, & Knowles, 1990), and the use of groups at the two extremes of a scale (Preacher, Rucker, MacCallum, & Nicewander, 2005). Although these discussions have primarily focused on the practice of using chopped data to perform analyses of data obtained from experimental designs (e.g., ANOVA), it has also been argued that applied research, where correlation and multiple regression are more popular, is not immune to this imprudent practice (Irwin & McClelland, 2001; McClelland & Judd, 1993).
