|
Abstract:
|
The bootstrap is one of the most efficient resampling methods to deal with small sample problems. However, the bootstrap mechanism in which the original data are resampled with replacement creates a new problem, dependence of the observations. To cope with this problem, the present study introduces an alternative resampling technique, resampling with random noises (RRN). The RRN resamples are randomly drawn from small Gaussian kernels estimated from the original data with random noises using automatically selected optimal bandwidth, which is methodologically different from the smooth bootstrap. The present study utilizes an empirical example and a Monte Carlo simulation to verify the effectiveness of RRN in reducing sampling errors; and the promising results suggest that RRN is a viable resampling method but easier to be implemented than the bootstrap.
|
