Piecewise distribution function estimation with Generalized Pareto for tail

Hello all,

I am trying to estimate the cumulative distribution function for a
single stock return time series. A piecewise estimation is composed of
three parts: parametric generalized Pareto (GP) for the lower tail (10%
of the observation), non-parametric kernel-smoothed interior (80% of the
observations), and GP for the upper tail (10%). I wonder if anyone has
clue about this in R.

The software of Matlab has a function called 'paretotails' in the
Econometrics Toolbox to do the estimation. On this site, a couple of old
messages were related but no clear answers were given.

Any help is greatly appreciated.


Edwin Sun