3. In your field of research, think of an established relationship between two variables at the same level of analysis

(e.g., a personality trait positively predicts individual task performance). For each variable, describe how you could aggregate it to compose a construct at the higher level of analysis. Now, consider the relationship between these two aggregated variables. Is this relationship meaningful in your field of research? Does the relationship tell you anything new beyond what you already know from the relationship between the two original variables

(prior to aggregation) at the lower level of analysis? Can you think of situations in which the original relationship at the lower level of analysis holds but the relationship linking the aggregated variables at the higher level of analysis does not hold, and vice versa? Based on your knowledge of the field of research, what are some of the similarities and differences that you think exist between the two relationships? This exercise is complex and in some ways goes beyond the basic issues discussed in this chapter. Hence, I suggest you perform this exercise one more time after you have read about some of the advanced issues pertaining to composition models in multilevel research (see Chan, 1998b; Hofmann, 2002; Kozlowski & Klein, 2000).