ESTIMASI MAKSIMUM LIKELIHOOD MELALUI ALGORITMA EKSPEKTASI MAKSIMISASI UNTUK MODEL REGRESI LINEAR DENGAN DATA HILANG; MAXIMUM LIKELIHOOD ESTIMATION VIA EXPECTATION MAXIMISATION ALGORTHM FOR LINEAR REGRESSION WITH MISSING DATA

HARIZAHAYU, Subanar

2014 | Disertasi | PROGRAM STUDI S2 MATEMATIKA

Data is one of the important points in every data analysis as it is impossible to conduct data analysis without data. The data used is expected to be good data. In fact, it is commonly found that the data doesn’t meet the expectation. Incommplete data causes the difficult in drawing the conclusion. If missing data are ignored, it causes the conclusion are bias or invalid. This research will use a linear regression model. Regression analysis is a statistical analysis is performed to model the relationship between Y (the dependent variable) and a categorical random variable X (the independent variable). For continuous variables for Y and discrete variables for X, assuming that all of the observed variables Y and X are some missing variables. The classification of missing data to be compared consists of three classifications, namely: MCAR, MAR, and MNAR. This discussion concludes with a case study on the estimated value of the missing data on the variable data presentation xerostomia using EM algorithm to calculate the maximum likelihood estimation (MLE) in a linear regression model with the three classifications of missing data.

Kata Kunci : linear regression, missing data, missing completely at random, missing at random, missing not at random, Lagrange multipliers, ekspectation maximization.