آیا بحث بیشتر در مورد سطوح تجزیه و تحلیل واقعا لازم است؟چه مواقعی چنین بحث های کافی است؟

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.

After reading the articles by Dickson, Resick and Hanges (2006-this issue), Peterson & Castro (2006-this issue), and Hanges & Dickson (2006-this issue), we were struck not so much by the differences between these authors' perspectives, as by the potential complementary nature of the authors' views and recommendations. Thus, the first part of our article focuses on the similarities among the prior three articles and develops a set of theoretical and empirical requirements necessary to address the multi-level theoretical and empirical issues raised by the GLOBE study. We then ask, Are these requirements, taken together, sufficient to address the theoretical questions of interest to all the authors? To evaluate their sufficiency, we introduce mainstream, well-known, multi-level information that has been available for more than 50 years. This information both adds to and informs the multi-level theory of the authors. In addition, this information, which shows the relationship among all the authors' ideas, allows for different theoretical positions to be tested using data from the GLOBE study and similarly well-designed multi-level studies. The same logic can also be applied to measurement. Based on this material, we present an analytic checklist for use by researchers who are doing multi-level studies similar to the GLOBE study. Finally, we show the compatibility of this approach to a variety of other approaches to multi-level research.

The combination of the articles in this issue offer significant advances over previous work. The first article (Dickson et al., 2006-this issue) provides insights into possible group effects based on the cross-level view. The second article (Peterson & Castro, 2006-this issue) raises questions about the adequacy of the methods in the first study and the failure of the first article to adequately test or allow for an emergent view. The third paper (Hanges & Dickson, 2006-this issue) elaborates on the need for additional analyses, including the use of structural equation modeling (SEM) after a consideration of the basic correlations. The present article suggests that all of these authors are correct in their own way, given their own starting points. We have suggested ways to test empirically the important alternatives raised in these papers. As exemplified by the approach to SEM described by Hanges & Dickson (2006-this issue), we hope that future research will consider all basic indicators, such as those listed in the checklist in Table 9, before attempting more advanced techniques, such as HLM or SEM, that require various assumptions. The ingenious design of the GLOBE project provides, in a cross-cultural study, an excellent opportunity to finally address the important issues that we have tried to point out here. Our checklist and the use of analyses that we described here should produce exciting new knowledge about leadership in different organizations and cultures.