Multilevel Analysis of Factors Affecting School Performance in Third Grade Spanish in Honduras

Abstract

In this paper, is made an analysis of school factors that affect performance in spanish third grade in Honduras, using data from Third Comparative and Explanatory Regional Study TERCE 2013, through a model of multilevel regression that allows performing analysis with estimates by levels; by the frequentist and Bayesian approaches. In the frequentist approach several models are made until the final model is obtained, and tests are made to verify the assumptions of homoscedasticity, orthogonality and normality, selecting the best model using AIC, BIC, and Log-Likelihood criterion. In the Bayesian approach, the models are compared using PSIS-LOO and WAIC criterion. Among the main findings, the variables highlighted with the greatest impact are: climate in the school classroom and recreation practices, the variables facilities and basic services of the school, indicate that better school characteristics have a positive and significant effect on student performance in Spanish.