Elmanani Simamora, Abil Mansyur, Eri Wydiastuti
Applying the ordinary least squares method in estimating the parameters of a linear regression model in the presence of heteroscedasticity becomes inefficient, even though the estimator is still unbiased. It is because the standard error estimate with ordinary least squares is no longer the smallest. As a result, conclusions from statistical tests can be misleading. One of the methods used to overcome this problem is the bootstrap method. There are three resampling techniques in the percentile bootstrap method that are considered to estimate the confidence interval of the heteroscedasticity linear regression model, namely residual bootstrap, wild bootstrap. The simulation results show that the wild bootstrap provides an average of the shorter confidence interval estimation length. © 2022 American Institute of Physics Inc.. All rights reserved.
Mathematics Department, Fmipa Universitas Negeri Medan, Jalan William Iskandar Ps. V, Medan, Indonesia
