BOOTSTRAP UNIFORM CONFIDENCE BANDS FOR A LOCAL LINEAR NONPARAMETRIC ESTIMATOR AND APPLICATIONS TO FINANCIAL RISK MANAGEMENT

Abstract

This paper considers the problem of bootstrapping a local linear estimator in conditional quantile estimation of a financial time series assuming independent and identically distributed errors. A nonparametric regression bootstrap generating process is estimated, then bootstrap confidence bands fitted to the quantile estimates. Under appropriate assumptions, the local linear bootstrap estimator is known to be consistent.