Elmanani Simamora, Abil Mansyur, Muliawan Firdaus, Rizki Habibi
High-performance concrete data has characteristics of non-normally distributed residuals and non-homogeneous residual variances. Violation of the assumption of non-normal residual distribution and non-constant residual variance will provide statistics or inferences that can be misleading. This study offers residual transformation through scaling and residual bootstrap methods. The statistical output of the simulation shows that the least squares method underestimates the three residual bootstrap methods. The classical residual bootstrap method experiences an insignificant increase in performance compared to the other two residual bootstrap methods. The student and standard residual bootstrap methods have almost similar statistical performance. However, the statistical performance of the standard residual bootstrap method is superior to other methods. © 2025 Author(s).
Departement of Mathematics, FMIPA Universitas Negeri Medan, Kota Medan, Indonesia
