| Abstract: | The use of the hierarchical linear modeling (HLM; A. S. Bryk & S. W. Raudenbush, 1992) statistical procedure has recently become more common. The author reviews HLM, discusses its advantages over ordinary least squares regression in modeling cross-level data, and demonstrates advantages to group researchers: more precise estimates of the relative strength of the relationships between variables at 2 or more levels of analysis; increased power due to distinguishing between random error and "error" attributed to between-group variance; clarity about if and why some group properties might affect the strength of bivariate individual-level relationships; random sampling only at the highest level of analysis, because levels nested within are assumed to be intercorrelated; and the choice of comparing individuals to the whole population or just those within the same groups. (PsycINFO Database Record (c) 2010 APA, all rights reserved) |
