[R-SIG-Finance] Boot samples from correlated time series.

[Noah Silverman]

Hi,

I'm looking at about 20 assets returns for the past 50 periods.  They
returns are correlated (some more than others.)

I'm interested in trying some monte carlo techniques for determining an
optimal portfolio.  The "usual" method I seen uses a multivariate norm
to sample possible data (rmvnorm function in R.)

The returns are definitely not normally distributed.  So, my goal is to
duplicate the same functionality of using a rmvnorm, but for unknown
distributions that are partially correlated.  It was suggested to me
that I might look at the boot function in R, however this appears to
generate a summary statistic which isn't what I want.

What I'd really like to do is simulate possible "alternate paths" for
the assets based on their historical data.  The paths are random, but
follow the same distribution as the real assets and the same
correlation.  I can't seem to find a way to do this.  Does anyone have
any suggestions?

Thanks,

--
Noah