[R-SIG-Finance] copula-based higher moment optimization

[Brian G. Peterson]

Alexander Moreno wrote:
> Is there any very quick and dirty way to do a copula-based optimization
> incorporating skewness using R packages, i.e. something that's close to
> being automated?  Any help would be appreciated.

Is there a particular paper that you're trying to emulate?

Here's the simple answer: there is no "automated" portfolio optimizer 
for higher moments in R.

Here's the complicated answer: there are many copula functions in R that 
could be used for a better estimate of the moments of your distribution, 
and those moments could be fed into an optimization routine.  There are 
also many non-copula methods in R for fitting non-normal distributions 
for a good estimate of the higher moments of the distribution (or, more 
appropriately, for a good estimate of the risk metrics at a particular 
quantile or confidence level).

Probably the "simplest" (most-automated) copula-based method for 
portfolio optimization in R that I've seen was done by some of Prof. 
Wuertz's graduate students, who were working on fitting a copula to the 
tail of the distribution to get a better estimate of Expected Shortfall, 
and then using that to build ES-optimal portfolios.  I assume that Prof. 
Wuertz's  team work will end up in RMetrics eventually, but I don't 
believe it's made it into the released code just yet: check the 
examples, tests, and documentation for the fCopula package to find out. 
You can also use a Cornish Fisher expansion to calculate a four-moment 
VaR (see package PerformanceAnalytics) and use that as the downside risk 
metric in your optimization.

Regards,

    - Brian