Elmanani Simamora, Abil Mansyur, Mukhtar, Muliawan Firdaus, Rizki Habibi
In the two-point asymmetric Bernoulli scheme, a logistic function is used to calibrate the probability against residual skewness directly. A basic linear regression model with right-skewed errors and a multiple model with left-skewed errors are then subjected to Monte Carlo evaluations. Three methods are used to create confidence intervals: the percentile method, the bias-corrected and accelerated method, and the bootstrap-t method. Two-point asymmetric Bernoulli can produce shorter intervals than classical weights while maintaining coverage probabilities at the nominal level, according to simulation results. The bias-corrected and accelerated method gives the best balance between speed and precision. The percentile method provides shorter intervals, and bootstrap-t usually offers more stable coverage. The simulation results show that the two-point asymmetric Bernoulli method is a more effective way to resample data, as it is more flexible and dependable, mainly when the data exhibits heteroscedasticity and asymmetric residual distributions. The two-point weights are simple, which is a positive aspect because it allows the method to circumvent the problems associated with the standard bootstrap method. Comparative studies on real data show that the generalised linear model’s interval length is wider and less stable than that of the two-point asymmetric Bernoulli. © 2026 by authors, all rights reserved.
Department of Mathematics, Universitas Negeri Medan, Medan, 20221, Indonesia
