Citation

Rose RA. J. Fam. Violence 2018; 33(2): 109-122.

Copyright

(Copyright © 2018, Holtzbrinck Springer Nature Publishing Group)

DOI

10.1007/s10896-017-9938-z

PMID

unavailable

Abstract

Family violence researchers often use an ecological perspective to describe persons nested within groups. Further, family violence researchers frequently investigate whether group characteristics impact individual outcomes. The theoretical orientation and research designs typically used therefore present opportunities to utilize multilevel modeling (MLM) for clustered designs. It is widely understood that MLM corrects standard errors for grouped data, though other approaches can address this issue. Importantly, MLM presents a structured approach to the examination of group differences in outcomes, group differences in the association between the characteristics of persons and these outcomes, and the explanation of group differences using group-level characteristics. This journal frequently receives studies that use MLM for clustered designs, and a set of analytical guidelines may assist authors in preparing such articles so as to properly implement and better leverage the power of MLM to advance family violence research. I describe MLM for the new user, providing guidance on estimation of these models in the context of two examples. In addition, for more experienced users of MLM, I argue for greater attention to between-group and compositional effects that may be prevalent in family violence research, and the opportunities they may raise for a better understanding of the complexities at the group level. In closing I discuss some extensions of MLM and place MLM in the context of research design, providing guidelines for designing, carrying out, and reporting findings from studies that use these methods.

Language: en

Keywords

Cluster randomized trial; Clustered data; Ecological theory; Family violence; Multilevel modeling