KULLBACK’S INFORMATION CRITERION CORRECTION (KICC) UNTUK SELEKSI MODEL REGRESI LINEAR MULTIVARIAT MULTIVARIATE REGRESSION MODEL SELECTION USING KICC (KULLBACK’S INFORMATION CRITERION CORRECTION)

SARAH, ITTA AGATHYA, Zulaela

2015 | Skripsi | FMIPA

Regression analysis is a statistical analysis that used to perform model relationship between dependent variabel and independen variables. If there are several dependen variables which are correlated and one or more independen variables, then it called multivariate linear regression analysis. In regression, model selection is a important part in order to determine which predictor (independen) variables that significantly influence the response (dependen) variable in the regression model. We already know some of the test criteria used for selection regression model , and most often we use is the criteria AIC ( Akaike Information Criterion ) . But if the AIC is used on a small sample , AIC will be biased , so that the test criteria established KICC ( Kullback 's Information Criterion Correction ) as a refinement of the AIC. KICC is an expectation of Kullback-Leibler Symmetric Divergence between probability density of candidate model and a true model.

Kata Kunci : multivariate linear regression; model selection, small samples; Kullback Leibler symmetric divergence.