The simultaneous consideration of between-participants and within-participants analyses in research on predictors of behaviours: The issue of dependence

Although there is wide acceptance among personality and social psychologists of the importance of performing both between‐participants and within‐participants analyses to obtain a more complete picture of the phenomena under investigation, such analyses are rare (Mishela, 1990). Research on the predictors of behaviour, particularly concerning variables such as attitude, subjective norm, affect, cognition, and intention provides an exception, where the results from within‐participants analyses are sometimes compared to the results from between‐participants analyses. These comparisons raise the issue of whether the two types of analyses are independent of each other (and whether they can be validly compared), which is the topic of the present paper. Although we show that there is dependence, which suggests that it is a bad idea to compare both kinds of analyses, we also show that the degree of dependence approaches zero as the number of participants and items increase. Thus, with a sufficiently large design, the degree of dependence is unimportant, and therefore is no obstacle to the simultaneous consideration of both within‐participants and between‐participants analyses. How large is large enough? A set of computer simulations suggests that 15 participants and 15 items is sufficient, though we provide data from which researchers can designate their own criteria. Copyright © 2004 John Wiley & Sons, Ltd.