Pemodelan Bayesian Model Averaging (BMA) Pada Kasus Pneumonia Balita

ABSTRACT Bayesian method is known as a better method than other methods because it combines the information from the sample data and the information from the previous distribution (prior). There are several methods in the Bayesian able to choose the best models involving uncertainty models and one of them is Bayesian Model Averaging (BMA). BMA is a method that can predict the best model based on the weighted average of all models. BMA goal is to combine model uncertainty in order to get the best model. The purpose of the study was to determine the linear regression model of the BMA in cases of pneumonia. Design research is applied research. The experiment was conducted in Situbondo in May-June 2014. Sampling was done by total sampling 0f 17 health centers throughout Situbondo. BMA results indicate that there were 27 models selected with the 5 best models from the 2048 model is formed. BMA Model was produced 9 significant variable predictor of the response variable. These variables were not smoke in the house, healthy household, exclusive breastfeeding, infants received vitamin A, DPT immunization coverage, low birth weight, malnutrition children, number of posyandu and toddler health services. Variables were not significant are clean and healthy living behavior and infant visits. Keyword linear regression, bayesian model averaging, pneumonia