The (public domain) package BurStFin is now available at repository: http://www.burns-stat.com/R This largely consists of functions that have been split off from the commercial software. The functionality includes: * estimation of a variance matrix as a statistical factor model (missing values are allowed -- even all missing for some assets). * estimation of a variance matrix by shrinking to the equal correlation matrix (Ledoit-Wolf). This also allows missing values. * Add a benchmark to a variance matrix * Transform a variance matrix to be relative to a benchmark. The 'tawny' package also has a function to estimate the shrinkage variance. There is a slight difference in the estimated shrinkage. I have circumstantial evidence that the BurStFin function is what Ledoit-Wolf does, but it would be nice if someone spent the time to sort that out. For those looking for research topics, the help files for the two variance estimation functions have some questions that we don't seem to know the answers to. BurStFin will eventually, at the appropriate juncture, in due course, in the fullness of time, at the end of the day appear on CRAN. Caution: if you have R version 2.10.0, you need to update to 2.10.1 in order to have the repository work. Windows binary and source are available, but MacOS is not. However, all the code is R code -- there is no C or Fortran. More details at: http://www.burns-stat.com/pages/public.html
Patrick Burns patrick at burns-stat.com http://www.burns-stat.com