Is more discussion about levels of analysis really necessary?When is such discussion sufficient?

This article begins by showing why most of the recommendations made in the three previous articles about the original GLOBE study are necessary given the authors' viable and insightful theoretical interests. It then demonstrates that the recommendations made in these articles, although necessary to follow, are not sufficient to allow the authors to conclude what they want to say. This article reveals what is missing, and why, and then adds several more recommendations. All of these recommendations are summarized with a checklist that itemizes what a researcher interested in a multiple levels of analysis theory needs to consider when focusing on bivariate correlations or multiple regression. This checklist is used to show which items are and are not included in the original article by Dickson [Dickson, M.W. (2006-this issue). Systematic variation in organizationally-shared cognitive prototypes of effective leadership based on organizational form. The Leadership Quarterly 17, 487−505]. The design of the GLOBE study is then examined, with the conclusion that it satisfies all of the items in the checklist. These same issues arise regardless of whether one prefers η2, ICC(1), ICC(2), rwg, or multi-level modeling (for example, structural equation modeling or hierarchical linear modeling) to analyze multi-level data such as that represented by the GLOBE study.