Hi Feirong,
It generally depends on two factors: (1) your sample size at your highest
level of analysis and (2) the degree of balance (or lack thereof) in your
data. If your sample size at your highest level of analysis is large and
the data are not too unbalanced, the empirical Bayes approach used in HLM
is ok. What this essentially means is that unbalanced data is not a
problem because empirical Bayes esimates "borrow strength" from
information contained in other units (e.g. teams) by taking into account
the estimates for other units and the characteristics the units share. If
your sample size at your highest level of analysis is small and the data
are unbalanced, a "fully Bayesian" approach is a better option. This also
has implications in regards to your centering choice. I attached a lecture
from Feifei Ye's HLM class at the University of Pittsburgh that discusses
this in a little bit more detail. Below are some books/articles that may
also be helpful:
Raudenbush, S. W. & Bryk, A. S. 2002. Hierarchical Linear Models:
Applications and Data Analysis Methods (2nd ed.). Thousand Oaks, CA:
Sage. - See Chapter 13
Hofmann, D. A. 1997. An overview of the logic and rationale of
hierarchical linear models. Journal of Management, 23(6): 723-744.
Hofmann, D. A. & Gavin, M. B. 1998. Centering decisions in hierarchical
linear models: Implications for research in organizations. Journal of
Management, 24: 623-641.
Kreft, I. G. G., de Leeuw, J., & Aiken, L. S. 1995. The effect of
different forms of centering in hierarchical linear models. Multivariate
Behavioral Research, 30(1): 1-21.
Kevin S. Cruz
Ph.D. Candidate in Organizational Behavior and Human Resources Management
University of Pittsburgh
Joseph M. Katz Graduate School of Business
247 Mervis Hall
Pittsburgh, PA 15260
Phone: (727) 515-1151
Fax: (412) 624-3633
E-mail:
kscruz@katz.pitt.edu