A nonparametric threshold model with application to zero returns

We propose a nonparametric censoring model for time series data. We propose an estimator of the censoring function based on extreme value regression. We obtain the pointwise distribution theory and suggest confidence intervals based on this theory. We use our model to explain the evolution of the frequency of zeros in stock index returns.

Keywords

censoring, extreme value theory, GARCH, index returns

Full Text (PDF format)

Published 1 January 2008