Information Systems researchers are often concerned with empirical questions spanning more than one level of analysis. For example, virtual teams research provides a good illustration because such teams are inherently hierarchical entities involving the situated nature of individuals within teams. Despite the importance of multilevel research questions to Information Systems research, the literature has yet to fully engage appropriate techniques for multilevel investigations. Using hierarchical linear modeling (HLM) as a statistical tool that can appropriately test cross-level relationships, we provide an illustration of the differences and advantages of using a multilevel technique over ordinary least squares (OLS) regression. Using data from a study of global virtual teams, we demonstrate that substantive research conclusions differ based on the use of HLM versus OLS regression. Using HLM, we find a significant relationship between individual level task liking and affective commitment; we also find a significant relationship between individual level task liking and satisfaction with the virtual team. When testing the moderating effects of team characteristics, we found a significant positive moderating effect of team work processes on the relationship between task liking and satisfaction. We conclude with recommendations for future research and provide a comparison of empirical techniques available for IS researchers testing relationships at single and multiple levels of analysis.
Short, Jeremy; Piccoli, Gabriele; Powell, Anne; and Ives, Blake
"Investigating Multilevel Relationships in Information Systems Research: An Application to Virtual Teams Research Using Hierarchial Linear Modeling,"
Journal of Information Technology Theory and Application (JITTA): Vol. 7:
3, Article 5.
Available at: https://aisel.aisnet.org/jitta/vol7/iss3/5