Hierarchical linear modeling (HLM) is a regression-based analysis that takes the hierarchical structure of the data into account. Hierarchically structured data is described as nested data where nesting represents groups of units clustered together, such as students within classrooms. The nested structure of the data violates the independence assumption of ordinary least squares (OLS) regression, because clusters of observations (e.g., students at a school) are not independent of each other. That is, students at any given school share an environment that potentially explains a portion of the variance in the model’s outcome.
The flexibility of HLM models is reflected in the variety of applications for which it is used. In addition to the study of individuals in organizations as described above, HLM models are used in agriculture (e.g., crops within plots), ecology (e. g., animals in herds), and business (companies within industries). There is also substantial application of HLM models for the study of longitudinal data where observations are nested within individuals. Longitudinal HLM models, sometimes described as growth curve models, treat time in a flexible manner that allows the modeling of non-linear and discontinuous change across time and accommodates uneven spacing of time points and unequal numbers of observations across individuals.
HLM models provide a framework that incorporates variables on each level of the model. For example, both student characteristics, such as age, and school characteristics, such as graduation rate, can be modeled. HLM models can be extended beyond the two levels that characterize the examples thus far. For example, students nested within schools are nested within school districts. In addition to purely hierarchical structures, there is a class of models called cross-classified models that allow units to be nested within more that one cluster where the clusters are not structurally related. For example, students could be nested within schools and churches, where there is no relationship between schools and churches.
HLM can be explained with the help of the above example in two steps. In the first step, separate analyses are conducted for every school taken under consideration with the help of the student level data. For example, the test scores of the students in some particular subjects could be regressed on the basis of the set of the student level predictor variables, like the student’s socio-economic status and a binary variable indicating the gender of the student.
In the second step, the regression parameters obtained from the first step of the analyses become the outcome variables of interest.
The underlying idea of Hierarchical linear modeling is that there involves separate analyses for each unit in a structure. There also exists certain complex Hierarchical linear models in Hierarchical linear modeling. In such types of models in Hierarchical linear modeling, the statistical analyses specified at each level are not only linear regression, but also certain higher level regression. In other words, HLM can also be called multi-level modeling.
In comparison to the classical type of regression, Hierarchical linear modeling is an improvement. Hierarchical linear modeling can be used for the purpose of prediction. It also can be used for the purpose of data reduction, and can be helpful for drawing out the causal inference.
The Hierarchical data structure has basically two types of categories according to Bryk and Raudenbush (1992). The first category of the data structure in Hierarchical linear modeling includes repeated measure types of data and the second category includes the meta-analytic type of data. The meta-analytic type of data is that type of data that deals with the large number of existing studies.
Hierarchical linear modeling becomes an issue or a problem when there is increased homogeneity within the hierarchical data structures. In other words, there is the problem of independence of observations. HLM violates this assumption of independence of the observations.
Hierarchical linear modeling also becomes a problem in the case of cross level data. A way of stabilizing this type of problem is to cumulate the data structures. Hierarchical linear modeling is a powerful tool that can be used by the researcher to achieve more appropriate analyses of monitoring data. It cannot compensate the deficiencies in the quality of data being collected by the researcher, which is similar to other types of analytical tools.
Hierarchical Linear Modeling Resources
Bryk, A. S., Raudenbush, S. W., & Congdon, R. H. (1996). Hierarchical linear and nonlinear modeling with the HLM/2L and HLM/3L Programs. Chicago: Scientific Software International. View
Goldstein, H. (1991). Nonlinear multilevel models with an application to discrete response data. Biometrika, 78(1), 45-51.
Goldstein, H. (1995). Multilevel statistical models (3rd ed.). Oxford, UK: Oxford University Press.
Hoffman, D. A., & Gavin, M. B. (1998). Centering decisions in hierarchical linear models: Implications for research organizations. Journal of Management, 24(5), 623-641.
Julian, M. W. (2001). The consequences of ignoring multilevel data structures in nonhierarchical covariance modeling. Structural Equation Modeling, 8(3), 325-352.
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.
Li, F., Duncan, T. E., Harmer, P., Acock, A., & Stoolmiller, M. (1998). Analyzing measurement models of latent variables through multilevel confirmatory factor analysis and hierarchical linear modeling approaches. Structural Equation Modeling, 5(3), 294-306.
Morris, C. N. (1995). Hierarchical models for educational data: An overview. Journal of Educational and Behavioral Statistics, 20(2), 190-200.
Nezlek, J. B., & Zyzniewski, L. E. (1998). Using hierarchical linear modeling to analyze grouped data. Group Dynamics: Theory, Research, and Practice, 2(4), 313-320.
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods (2nd ed.). Thousand Oaks, CA: Sage Publications. Hierarchical Linear Models: Applications and Data Analysis Methods (Advanced Quantitative Techniques in the Social Sciences)
Singer, J. (1998). Using SAS Proc Mixed to fit multilevel models, hierarchical models, and individual growth curves. Journal of Educational and Behavioral Statistics, 24(4), 323-355.
Sullivan, L. M., Dukes, K. A., & Losina, E. (2004). Tutorial in biostatistics: An introduction to hierarchical linear modelling. In R. B. D’Agostino (Ed.), Tutorials in biostatistics. Volume 2: Statistical modeling of complex medical data (pp. 35-126). Hoboken, NJ: John Wiley & Sons.
Wendorf, C. A. (2002). Comparisons of structural equation modeling and hierarchical linear modeling approaches to couples’ data. Structural Equation Modeling, 9(1), 126-140.
Wong, G. Y., & Mason, W. M. (1985). The hierarchical logistic regression model for multilevel analysis. Journal of the American Statistical Association, 80, 513-524.
Related Pages:
