论文信息 - Bootstrap selection of bandwidth and confidence bands for nonparametric regression

Bootstrap selection of bandwidth and confidence bands for nonparametric regression

A bootstrap method is developed to estimate the average squared error of a kernel based nonparametric regression estimator for a given bandwidth. This estimated average squared error is then minimised over the bandwidth to produce a regression estimate. Locally adaptive smoothing and simultaneous confidence bands may be obtained from this bootstrap method.

Julian J. Faraway

[1] D. M. Titterington,et al. On confidence bands in nonparametric density estimation and regression , 1988 .

[2] H. Müller,et al. Estimating regression functions and their derivatives by the kernel method , 1984 .

[3] W. Härdle,et al. Bootstrapping in Nonparametric Regression: Local Adaptive Smoothing and Confidence Bands , 1988 .

[4] R. Stine. Bootstrap Prediction Intervals for Regression , 1985 .

[5] H. Müller,et al. Kernel estimation of regression functions , 1979 .

[6] J. Faraway,et al. Bootstrap choice of bandwidth for density estimation , 1990 .

[7] W. J. Padgett,et al. Smooth nonparametric quantile estimation under censoring: simulations and bootstrap methods , 1986 .

[8] Peter Hall,et al. Using the bootstrap to estimate mean squared error and select smoothing parameter in nonparametric problems , 1990 .

[9] W. Härdle,et al. How Far are Automatically Chosen Regression Smoothing Parameters from their Optimum , 1988 .