[R-SIG-Finance] [R-sig-finance] Any suitable backtest functions?

[Vijay Vaidyanathan]

On Jan 21, 2009, at 9:08 AM, R at Nabble wrote:
>
> After learning from replies of the list, I think maybe I shoud not use
> 'backtest' here.

I don't know if this will help, but one approach to portfolio  
simulation is: to assemble portfolios at regular intervals in time by  
using a certain selection criterion, and then simulate the behaviour  
of that portfolio over time. Youi'll need to specify how to weight  
the portfolio and how long to revisit the weighting and reselection  
decisions etc. This is the model that is used in a number of academic  
studies.

A couple of years ago, I implemented this model, and wrote a  
portfolio simulation package in R. I called it PAST, (Portfolio  
Attribution and Simulation Toolkit) and it was designed to be  
database independent. Unfortunately, it isn't in CRAN yet, but I've  
spent the last couple of months cleaning it up and bundling it as an  
R package.

The major outstanding issue is that It still lacks decent  
documentation, which I am working on and expect to get done in the  
next month or two, after which I hope to submit it to CRAN. In the  
meanwhile, if you can't wait for a month or so, just drop me an email  
message off-list and, I'll send you a link to it on the web so you  
can play with it.

Also, if you are in the San Francisco Bay Area, I believe that Dave  
Stewart will be giving a talk at the monthly meeting of the Silicon  
Valley SI Pro User Group of the American Association of Individual  
Investors on Feb 2 (at the Saratoga Public Library, 6:30 pm). He'll  
talk about using the software, including helping you getting it  
installed and running some basic simulations. Unfortunately, I will  
be traveling and won't be at the meeting, but you'll be in good hands  
with Dave!

- Vijay
======