Understanding the structure of data when planning for analysis: application of Hierachical Linear Models

Abstract

Human beings and other living creatures tend to exist within organisational structures, such as families, schools, and business organisations. In an educational system, for example, students exist within a hierarchical social structure that can include classroom, grade level, school, school district and country. Data obtained from such social structures are hierarchical. It is critical that social scientists understand the structure of the data because it dictates the statistical techniques to be used for analysis and interpretation. For example, analysing hierarchical data using the conventional General Linear Models (GLMs) may result in inaccurate inferences being drawn from the data. A thorough understanding of the data in terms of structure, type of variables and relationships being investigated needs no further emphasis. Statistically valid inferences are drawn from data that have been carefully collected and subjected to the appropriate statistical techniques. Attention should also be paid to the underlying assumptions of a particular statistical technique. Use of Hierarchical Linear Models (HLMs) in analysing social research has several advantages. The problem of unit of analysis is avoided and data are no longer aggregated or disaggregated resulting in accurate and reliable estimation of each level effects. Furthermore, all estimated effects are adjusted for individual level and group level influence on the outcome variable. The only drawback of applying HLMs is that this requires an advanced level of sophistication in statistics.