Modified geographically weighted regression to accommodate the spatial lag of predictor in the model (case study of stunting data in East Nusa Tenggara Province 2021)

• Main Authors: Abdila, Naufal Shela, Fitriani, Rahma, Soehono, Loekito Adi
• Format: Conference Proceeding
• Language: English
• Subjects: Autocorrelation; Education; Heterogeneity; Homogeneity; Households; Hypotheses; Null hypothesis; Parameter estimation; Regional development; Regression analysis; Rejection
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0166505

Geographically Weighted Regression (GWR) is a regression analysis which is developed specifically to estimate the magnitude and the direction of causal relationship between one response variable and predictor variable(s), by taking account the location heterogeneity. The model parameters are estimated differently for each location to take account the spatial heterogeneity. The problem of stunting in Indonesia, especially in East Nusa Tenggara, indicates the presence of spatial heterogeneity. The situation is confirmed by the rejection of null hypothesis of spatial homogeneity, in the Breusch Pagan test (9,026) for the regional percentage of stunting case. Therefore, GWR is an appropriate model to capture the regional percentage of stunting case, as a function of several predictors. Among the proposed predictors, several regional economic indicators area used (i.e., GRDP growth rate, real expenditure per capita). Those are typically variables with spatial autocorrelation, which is confirmed by the rejection of spatial dependence hypothesis in the Moran’s I test (7,66×10−8). Therefore, this study is aiming at modifying the GWR, to accommodate the spatial lag of predictors in the model. The modified model is the used to analyze factors which are mostly determined the percentage of stunting case in East Nusa Tenggara in 2021. Fixed gaussian spatial weight and rook contiguity spatial weight is used to capture the spatial heterogeneity and spatial autocorrelation respectively. In general, the variables that affect the prevalence of stunting in East Nusa Tenggara Province are the percentage of households that do not use defecation facilities, percentage of mothers who do not attend formal education, and GRDP growth rate. The variable that most increases the prevalence of stunting in East Nusa Tenggara Province is the variable percentage of mothers who do not attend formal education.