Estimasi Model Linier Tergeneralisasi Gaussian Berdasarkan Maximum Likelihood Estimator Dengan Menggunakan Algoritma Fisher Scoring
Abstract: Linier classic model
is linier model which is the response variable has normally distribution. When
the distribution of the response variable is the exponential family then the
distribution is called generalized linier model. In this final report used the
gaussian distribution. The aims of this final report are to obtain parameters
estimators of generalized linier modelgaussian based on maximum likelihood
estimation with residual deviance test. Maximum likelihood method used to find
parameter estimator with maximize the parameter, while residual deviance test
is used to measure how big a group of predictor variable to response variable.
Since the result implicit estimate, Fisher Scoring Algorithm used in this final
report. The Fisher Scoring Algorithm is Newton Raphson development which
replaces the hessian matrix by its expectation.
The final of applying generalized linier model-gaussian through S-Pluss
2000 Software on economic growth of east java on January 2007 until December
2010. Response variable used in this final report is economic growth of East
Java on January 2007 until December 2010 and predictor variables are income
from export, agriculture and import. Based on the data analysis we get the estimate
of income export is -3e-005, income agriculture is 0.02418 and income
agriculture is 0.0202. Based on the residual deviance test, it means that economic growth of east
java will progressively 61% with increasing income from export, agriculture and
import. In the suitability test the prediction is fairly well.
Keywords: Generalized linear model, Gaussian
Distribution, Maximum likelihood, Residual DevianceTest
Penulis: Zahrotul Ummah,
Suliyanto & Sediono
Kode Jurnal: jpmatematikadd130071
