[R-sig-Finance] Confidence intervals for spread returns

[David Kane]

Krishna Kumar writes:
 > are the portfolio weights found  using simulation. i.e. You simulate the 
 > underlying asset returns and then optimize
 > on the simulated paths?. Bootstrapping seems like a natural way to get 
 > the conf.intervals.

The backtest package, in its current form, is way simpler than
this. All we are doing (for now) is forming quintile (or decile or
whatever) portfolios and reporting the spread returns. We also allow
the user to stratify these by, for example, sector. The reliability of
the spread returns then depends, obviously, on the number of
observations in each sector. The confidence interval is just supposed
to provide a rough guide which takes this effect into account.

A bootstrap approach seems useful and productive, but we will probably
not be able to integrate it in the first release.

Dave