I just committed revision 47 of LSPM on R-forge, which includes
several bug fixes. Please wait a day or two for the packages to be
built by the R-forge server before running:
install.packages("LSPM",repos="http://r-forge.r-project.org")
Or you can always checkout the source and build the package yourself.
Best,
--
Joshua Ulrich | FOSS Trading: www.fosstrading.com
On Wed, Nov 17, 2010 at 9:14 AM, RVince <rvin...@hotmail.com> wrote:
> Josh,
>
> You are the man. Is this the same as I had on the memory stick to the
> guys in Tampa this past weekend/ -Ralph Vince
>
It's the same, except for one small bug fix. If you recall, I
mentioned the RNG wasn't running on the second of two consecutive
calls to optimalf with a drawdown constraint.
I thought it was a DEoptim issue. It turned out to be that the
probDrawdown C function always got/set the RNG state but only called a
RNG function when using the sampling shortcut, so the state wasn't
being advanced when we were *not* using the sampling shortcut.
R-forge was having issues building packages earlier this week (many
packages, not just LSPM), but it looks like all is well now. Please
let me know if you're still having issues.
Best,
--
Joshua Ulrich | FOSS Trading: www.fosstrading.com
It looks like you need to update DEoptim as well. LSPM used to
provide a copy of DEoptim, but it now relies on the DEoptim package.
Try:
install.packages("DEoptim")
HTH,
--
Joshua Ulrich | FOSS Trading: www.fosstrading.com
You need to change this line:
DEctrl <- list(NP=np,itermax=imax, CR=crossover, refresh=1)
to this:
DEctrl <- list(NP=np,itermax=imax, CR=crossover, trace=1)
The argument was re-named from "refresh" to "trace" from DEoptim_1.x
to DEoptim_2.x. Sorry for the trouble.
Best,
--
Joshua Ulrich | FOSS Trading: www.fosstrading.com
Sometimes install.packages doesn't work, so you should check the sites above.