Analyzing SEBD using structural equation modeling and multilevel modeling

Abstract

Multilevel modeling (MLM) is a statistical tool used to analyse clustered data structures where observations are nested within higher levels of classification and where processes occurring at a higher level of analysis influence the characteristics and processes occurring at a lower level. Multilevel models measure the variation at each level of nesting by including an error term at each level. Exploratory Factor Analysis (EFA) explores the data structure to identify latent variables. Additionally, Confirmatory Factor Analysis (CFA) confirms the latent structure whilst Structural Equation Modeling (SEM) assesses the relationships between the latent variables. SEM accommodates both exploratory and confirmatory modeling. This implies that these models are suited to both theory testing and theory development. Multilevel Structural Equation Modeling (MSEM) is a statistical technique that combines Multilevel modeling with Confirmatory Factor Analysis/ Structural Equation Modeling. Multilevel Structural Equation Modeling is appropriate to analyse datasets that include both latent variables and a nested hierarchical structure. Consequently, MSEM enables the researcher to fit a CFA/SEM model at every nesting level of the hierarchical structure. MLM, EFA, CFA, SEM and MSEM techniques are used in this thesis to analyse a dataset which is related to the students’ Social Emotional Behaviour Difficulties (SEBD). The Strengths and Difficulties Questionnaire (developed by Goodman (2005)) consists of 25 items, which assess four difficulty subscales (Hyperactivity, Conduct, Emotion and Peer problems) and a Prosocial subscale. In this thesis, the rating scores obtained from the Teachers’ SDQ were analysed. Each item is measured on a 3-point scale ranging from 0 to 2, where the score for each subscale ranges from 0 to 10 and the total difficulty score ranges from 0 to 40. The total difficulty score excludes the Prosocial subscale. The teachers, which were randomly selected from 110 schools, provided the rating scores on these 25 items for 5,200 students selected from 1,326 classrooms. MLM is fitted using the facilities of GLLAMM within STATA (Version 11), assuming a three-level nesting structure, where students are nested within classrooms which in turn are nested within schools. Additionally, a part of EFA is conducted using the facilities of SPSS (Version 22). Moreover, EFA, CFA and SEM are carried out using the facilities of LISREL (Version 8.80). Furthermore, MSEM is conducted using the facilities of Mplus (Version 6) by fitting a two-level CFA and a two-level SEM model.