Multilevel analysis is important in analysing hierarchically structured data. It aims not only to study the variations in the group but also the variation between groups. Meanwhile, in simple regression analysis, it involves data at one level only (group presence is negligible). Often we look at the research on hierarchical data (particularly in education studies) ignores the presence of level in hierarchical structure (e.g., class, schools) whereas each level also contribute a variation in the data. Therefore, this article will discuss the importance of doing multilevel analysis compared with simple regression analysis. Procedure for construction of multilevel estimation model is also shows in stages to observe the effects changing of the variations contributed by the group. Three types of multilevel models were compared with simple regression model, OLS and founds that multilevel model was better than the OLS model. Results also showed that the multilevel model with the inclusion of the level-2 variables gives the least model error and level-2 variance error. A simple example is used to facilitate the understanding of the fundamental in the construction of multilevel model.

You do not currently have access to this content.