Pemodelan Semiparametrik Geographical Weighted Logistic Regression pada Data Kemiskinan di Provinsi Sulawesi Selatan Tahun 2017

The level of poverty in a Regency/city in South Sulawesi in 2017 is different. The grouping of poverty status can be done based on the value of the HeadCount Index (HCI) of South Sulawesi. Factors affecting poverty will differ for each area being observed. The statistical modeling method developed for data analysis by taking into account the location factor is semiparametric Geographical Weighted Logistic Regression (GWLR). The GWLR semiparametric Model consists of parameters that are affected by the location and not affected by the location. The parameter estimator of the GWLR semiparametric model used in this research was obtained using the maximum method likelihood estimation. The result of a semiparametric model of GWLR each district/city in South Sulawesi in 2017 has the value Estimator parameter for global parameters is the same value for each location, namely, a3 = 0.1724, a4 = 0.0204, and a6 = 0.0261 whereas the parameter estimator for local parameters has different values so that GWLR semiparametric model of each district/city.

Evaluating class and school effects on the joint student achievements in different subjects: a bivariate semiparametric model with random coefficients

This paper proposes an innovative statistical method to measure the impact of the class/school on student achievements in multiple subjects. We propose a semiparametric model for a bivariate response variable with random coefficients, that are assumed to follow a discrete distribution with an unknown number of support points, together with an Expectation-Maximization algorithm—called BSPEM algorithm—to estimate its parameters. In the case study, we apply the BSPEM algorithm to data about Italian middle schools, considering students nested within classes, and we identify subpopulations of classes, standing on their effects on student achievements in reading and mathematics. The proposed model is extremely informative in exploring the correlation between multiple class effects, which are typical of the educational production function. The estimated class effects on reading and mathematics student achievements are then explained in terms of various class and school level characteristics selected by means of a LASSO regression.