The S-type ridge regression estimators with minimum mean squared error of the model and standard errors of the regressors
Makaleler > Tam Makale · 2026
Yazarlar: Filiz KARADAĞ, Hakan Savaş SAZAK
Journal of Statistical Computation and Simulation, Uluslararası, Wos, 2026, cilt 96, ss 3452-3476
DOI: 10.1080/00949655.2026.2669772
Özet
{Simulation is used to demonstrate a significant improvement in Ridge Regression.}In this study, we developed the S-type ridge regression estimators as a new solution in which multicollinearity and outliers exist. We showed that they have the minimum mean squared error (MSE) of the model and the standard errors of the regressors in all situations and the maximum r squared in almost all situations through extensive simulations. The simulation results also showed that they have overwhelmingly superior performance compared to the other ridge regression estimators in terms of the MSE, standard error and r squared efficiencies and even better under the Dixon’s outlier model, which shows that they are also robust. At the end of the study, we presented the results of four real-life data applications confirming the superior performance of the S-type ridge regression estimators, which yielded the lowest MSE and standard error, and the highest r squared values among all the estimators mentioned here.