METODE ROBUST BOOTSTRAP LEAST TRIMMED SQUARE PADA ANALISIS REGRESI LINEAR BERGANDA UNTUK KASUS PENCILAN DALAM SAMPEL KECIL
Abstract
One of methods to estimate parameters of multiple linear regression model at data contaminate outliers are Robust Least Trimmed Square (RLTS). RLTS estimates parameters of multiple linear regression model with trimming residual to get minimum of objective function. Then this method combined with bootstrap become Robust Bootstrap Least Trimmed Square (RBLTS) to solve small sample.
In this minithesis is discussed about the identification of outliers with RLTS estimator and estimates parameters of model with RBLTS. To shows RBLTS method resistance of outliers in small sample used secondary and simulation data with the number of observations are 10, 25, 30, 50 and the number of different resampling. The result showed that RBLTS method was better than RLTS because RBLTS has large value coefficient of determination and small residual sum of squares.
Description
Keywords
outliers, bootstrap, RLTS