Multilevel Modeling: Clarifying Issues of Concern

Abstract

When using Hierarchical Linear Modeling (HLM) to analyze complex nested data, it is important to consider issues that affect the interpretation of the HLM outcomes. Alternative methods of accounting for the variance within the nested structures need to be considered if they better fit the research question of interest. One alternative method to HLM is Multiple Linear Regression (MLR)-Ordinary Least Squares solutions with person vectors. This study compares a number of sets of data that reflect interaction questions as well as nested designs. More specifically, eight issues that need to be considered when using HLM are discussed. These issues are: 1) advantages of HLM; 2 & 3) person vectors as it relates to nesting; 4) centering; 5) picking the appropriate error terms for fixed, random, and mixed effects; 6) understanding interaction and how it is tested; 7) sample size related to first and second level models; and 8) comparing similarities and differences between HLM and MLR with person vectors.