Analysing SEBD using generalize linear multilevel mixed models

Abstract

Multilevel modeling is a statistical technique that facilitates the analysis of hierarchically structured data where observations are nested within higher levels of classification in which processes occurring at a higher level of analysis influence the characteristics and processes occurring at a lower level. An advantage of using multilevel modeling is the ability of this technique to separately estimate the predictive effects of an individual explanatory variable and its group-level mean. This modelling technique is used to relate social, emotional and behavioural difficulties (SEBD) scores of 1314 primary and secondary students nested in 196 classes, which in turn are nested in 42 schools. The nested structure of the data makes it possible to use multilevel models. Every student has a unique mix of demographic, family background, class and school environments level factors, which will be the model predictors. Since the score distributions are right skewed, the fitted multilevel generalized linear mixed models assume a gamma distribution and a reciprocal link function. These models accommodate both fixed and random effects and combine the theoretical framework of generalized linear models and multilevel models. These models allow the investigation of the basic individual and socio-cultural factors that may lead to social, emotional and behavioural difficulties and examine how these problems differ between classes and schools.