The researchers developed a new method called RLSRDSM to improve regression analysis when data has outliers and varying error variances. They also introduced WRLSRDSM to handle non-normal errors and heteroscedasticity. These methods outperformed the 2D-RDLS procedure in simulations with different sample sizes, outlier levels, and error distributions. The RLSRDSM and WRLSRDSM estimates showed robustness in handling both continuous and categorical variables in regression models.