Risk Factors for Malaria in Thailand Using Generalized Estimating Equation (GEE) and Generalized Linear Mixed Model (GLMM)

Abstract

         The objective of this research involved the identification of risk factors for malaria in Thailand using suitable statistical models. The Generalized Estimating Equation (GEE) and the Generalized Linear Mixed Model (GLMM) were employed. For both the GEE and the GLMM, the dependent variable with a Poisson distribution was compared with a Negative Binomial distribution. Secondary and provincial-level data used in this research were from several sources. The dependent variable was the number of people with malaria in 2007, and the independent variables were region, border, season, forest area, rain, temperature, and income. Suitable models were judged by the value of mean deviance (for GEE) and mean generalized chi-square (for GLMM). The study found that for both the GEE and the GLMM, the dependent variable with Negative Binomial distribution was more suitable than was the Poisson distribution. For the GEE, the population- averaged model, region (East and South), border (Myanmar, Malaysia, Cambodia, and Laos), season (May-Jul, Nov-Jan, and Aug-Oct), temperature, and rain significantly impacted the malaria incidence rate. For the GLMM, subject-specific model, factors that significantly related to the malaria incidence rate were region (East, South, and North-East), border (Myanmar, Malaysia, and Cambodia), and season (May-Jul and Nov-Jan).

Key words: malaria, Generalized Estimating Equation (GEE), Generalized Linear Mixed Model (GLMM), Poisson, Negative Binomial