Abstract:
The purpose of this study is to compare the estimation methods of linear regression model for data from lognormal distribution under random right-censoring from beta distribution which are Ordinary Least Squares Method (OLS), Chatterjee and McLeish Method (CM), Maximum Likelihood Estimation using the EM algorithm Method (MLE_EM), Maximum Likelihood Estimation Method using the EM algorithm with mean-adjusted data (MLE_EM_MEAN), and Maximum Likelihood Estimation Method using the EM algorithm with median-adjusted data (MLE_EM_MED) by generating data 2187 scenarios. The results present that 1) OLS and CM are the most efficient methods if the sample size is small or moderate (n=30,50) with small censoring proportion (r1=10) varies by ratio of variance of independent variables to variance of error. 2) If the sample size is moderate (n=50) with moderate or large censoring proportion (r1=20,30) or large (n=100) the most efficient methods are separated by timing of data collecting point. When data is collected in early, middle, and end of time, the most efficient method is MLE_EM_MED, MLE_EM family, and MLE_EM and MLE_EM_MEAN, respectively. 3) The efficiency of all methods increases when sample size, random censoring ratio, ratio of variance of independent variables to variance of error increases or censoring proportion decreases.