Multivariate regression library for C++ under Windows

Discussion in 'Automated Trading' started by lolatency, Jul 3, 2009.

  1. What is the easiest and quickest way for me to get the ability to do multivariate regression and have access to the usual t and F statistics, and all the usual good stuff. Do I have any real open source options that will work for me under Windows?

    R embedded is proving to be unwieldy.
     
  2. rosy2

    rosy2

  3. Splat

    Splat

    I would bet that matlab has an extensive regression toolbox.. It also lets u use C++ code.
     
  4. Corey

    Corey

    I second the recommendation for GSL.
     
  5. +1 for GSL.

    BTW, didn't you ask this question on another board recently (Brian?) ?
     
  6. I might as well be Brian if it gets me answers.

    -Brian
     
  7. What have you tried w.r.t R?
    I think it's best to get R work with C++/Java...
     
  8. I compiled R.dll, included all the headers, then used the Rcpp and RInside classes. Then I wrote some wrappers to let me manage things like regressions and such. It works fine, until I compile it as a DLL and import into a third party process. Stability goes wonky then.

    But also, I don't like the way it's structured -- parsing text and constructing vectors -- it's tedious. Do you have a better way?

    The DCOM version was a hassle because you have to run the GUI, the proxy interface, and the DCOM server. I found that to be impractical also.
     
  9. How about other methods of connecting R to C++/Java? Say socket?

    I think the best way is to call R functions from within C++/Java.

    Why is DCOm bad?

    I think it's important to get R working with C++/Java so you could leverage other functionalities of R.
     
    #10     Jul 4, 2009