LIFE EXPECTANCY MODELING USING MODIFIED SPATIAL AUTOREGRESSIVE MODEL

Hasbi Yasin (Department of Statistics, Faculty of Sciences and Mathematics, Diponegoro University)
Budi Warsito (Department of Statistics, Faculty of Sciences and Mathematics, Diponegoro University)
Arief Rachman Hakim (Department of Statistics, Faculty of Sciences and Mathematics, Diponegoro University)
Rahmasari Nur Azizah (Data Science Institute, I-Biostat, Hasselt University Belgium)

Media Statistika
Vol 15, No 1 (2022): Media Statistika

The presence of outliers will affect the parameter estimation results and model accuracy. It also occurs in the spatial regression model, especially the Spatial Autoregressive (SAR) model. Spatial Autoregressive (SAR) is a regression model where spatial effects are attached to the dependent variable. Removing outliers in the analysis will eliminate the necessary information. Therefore, the solution offered is to modify the SAR model, especially by giving special treatment to observations that have potentially become outliers. This study develops to modeling the life expectancy data in Central Java Province using a modified spatial autoregressive model with the Mean-Shift Outlier Model (MSOM) approach. Outliers are detected using the MSOM method. Then the result is used as the basis for modifying the SAR model. This modification, in principle, will reduce or increase the average of the observed data indicated as outliers. The results show that the modified model can improve the model accuracy compared to the original SAR model. It can be proved by the increased coefficient of determination and decreasing the Akaike Information Criterion (AIC) value of the modified model. In addition, the modified model can improve the skewness and kurtosis values of the residuals getting closer to the Normal distribution.