Weighted Split Sample Bootstrap for Regression Models with High Dimensional Data

Mazni Mohamad, Norazan Mohamed Ramli, Nor Azura Md Ghani
2016 Indian Journal of Science and Technology  
The bootstrap technology introduced by [1] has wide applications, particularly, in regression analysis. It has been used by researchers to construct empirical distributions for estimates of the regression coefficients. In case of outliers in the data, the classical bootstrap procedure fails to give us fine results even if robust regression estimates are used. In this paper we introduced a new bootstrap procedure, called "split sample bootstrap" to handle outliers. The proposed bootstrap
more » ... e gives bootstrap estimates having smaller bootstrap estimates of the standard errors and as a result, we get narrow confidence intervals of the estimates.
doi:10.17485/ijst/2016/v9i28/97789 fatcat:37ceserxi5en7modd6uii7jns4