Bootstrapping a consistent nonparametric goodness-of-fit test

Abstract

In this paper, we employ the parametric bootstrap to approximate the finite sample distribution of a goodness-of-fit test statistic in Fan (1994). We show that the proposed bootstrap procedure works in that the bootstrap distribution conditional on the random sample tends to the asymptotic distribution of the test statistic in probability. A simulation study demonstrates that the bootstrap approximation works extremely well in small samples with only 25 observations and is very robust to the value of the smoothing parameter in the kernel density estimation.

Keywords:

• Bootstrap
• Goodness-Oe-Fit Test
• Kernel Density Estimaition
• Size
• Power