A0850
Title: Bootstrapping generalized empirical likelihood with many weak moment conditions
Authors: Wenjie Wang - Hiroshima University (Japan) [presenting]
Abstract: The main contribution is to theoretically analyze the application of bootstrap methods to generalized empirical likelihood (GEL) estimation when the available moment conditions may be weak and the number of moment conditions goes to infinity with the sample size. We demonstrate that previous nonparametric i.i.d. and efficient bootstrap procedures cannot consistently estimate the distribution of GEL estimators under many weak moment conditions. The primary reason is that these bootstrap procedures fail to capture the overall moment strength in the sample adequately. Such a bootstrap failure implies that the widely used inference approaches such as the bootstrap standard error and the percentile type method are invalid under many weak moments. We also find that the efficient bootstrap is asymptotically less distorted than the nonparametric i.i.d. bootstrap under the current context. Finally, we discuss modified bootstrap procedures that provide a valid distributional approximation.
