REGRESI KERNEL UNTUK DATA LONGITUDINAL; KERNEL LONGITUDINAL REGRESSION

TOTO HERMAWAN, Sri Haryatmi

2015 | Disertasi | PROGRAM STUDI S2 MATEMATIKA

Longitudinal data is the data obtained by observations made of n mutually independent objects with each object observed repeatedly in different time periods and between observations in the same object is dependent. Smoothing technique used in estimating the nonparametric regression models to longitudinal data is Local Polynomial Kernel estimator that does not ignore the correlation between observa- tions within the same object. Local Polynomial Kernel estimator can be obtained by minimizing the Weighted Least Square (WLS) and complete the Generalized Esti- mating Equation similarities (GEE). Then for the selection of the optimal bandwidth using Cross Validation (CV) and the application of data created using software pro- grams R.The data used in this research is secondary data obtained from IHC Raflesia Babadan KD II Jaranan, Banguntapan, Bantul with the response variable is infant growth and predictor variables is age.Based on the results of the application of the model is that the value of the optimal bandwidth is 1 and the order of the polyno- mial 1 when CV minimum of 42, 22083 so that the results obtained by the model estimation MSE of 7, 10795.

Kata Kunci : Longitudinal Data, Nonparametric Regression, Bandwidth, Generalised Estimating Equation (GEE), Weighted Least Square (WLS), Local linear estimator, program R