Using Hierarchical Linear Modeling to Study Social Contexts:The Case of School Effects HLM AND THE SOCIAL CONTEXT OF SCHOOLELES
Valerie E. Lee
School of Education
University of Michigan
EDUCATIONAL PSYCHOLOGIST, 35(2), 125–141
The author explains the logic of the paper in a very fluent way who describes the background information in detail that why and what school effects should be considered.
This research was typically formulated in a two-stage process. The first stage identified schools that are particularly effective for low-socioeconomic status (SES) children. In the second stage, researchers searched for characteristics that were common among the schools identified as effective.
First, the author made a brief explanation of the HLM methodology, especially of exploring the social context of schools.
The second and third sections describe two studies that use HLM to explore school context effects
In the last section, the author revisits the issue of contextual effects, again through the school effects lens, to argue for the usefulness (indeed the necessity) of this approach to explore how educational contexts influence students.
These studies involve a search for statistical associations between school factors, on one hand, and variables measured on students on the other.
Three major difficulties commonly occur in the analysis of multilevel data and questions with single-level methods such as OLS regression, ANOVA, or structural equation modeling: aggregation bias, misestimated standard errors, and heterogeneity of regression. Aggregation bias can occur when a variable takes on different meanings and, therefore, has different effects at different levels of aggregation.
A second difficulty concerns the estimation of the standard errors used for statistical testing. With multilevel data, misestimated standard errors can occur when researchers treat individual cases as though they are independent (a standard assumption of OLS regression methods) when they are not.
A third difficulty concerns heterogeneity of regression slopes. That is, relations between characteristics of students (such as race, ethnicity, SES) and academic achievement may vary across schools and may be functions of group-level variables.
The paper gives us some perspective of why we should use HLM methodology and which variables should be chosen to analyze and the examples used in demonstrating is very useful for us to understand.