best language for quantiative analysis

Discussion in 'Automated Trading' started by Batman28, Jun 13, 2006.

which language..

  1. Matlab

    17 vote(s)
    56.7%
  2. Maple

    0 vote(s)
    0.0%
  3. Mathematica

    1 vote(s)
    3.3%
  4. oxmetrics - the whole family (pcgive etc.)

    0 vote(s)
    0.0%
  5. Gauss

    0 vote(s)
    0.0%
  6. Eviews

    3 vote(s)
    10.0%
  7. R

    8 vote(s)
    26.7%
  8. S plus (S+)

    1 vote(s)
    3.3%
  9. RATS & CATS

    0 vote(s)
    0.0%
  10. TSP

    0 vote(s)
    0.0%
  1. we more or less agreed that Python and C++ are the best languages for over-all trading system design.

    but what package is best for quantiative analysis/Econometrics/forecasting/modelling? ..e.g. GARCH, cointegration, forecasting models etc.. most are very compatable with c++ and excel. eitherway, which is the best on the following bases:

    1. usability, accebility
    2. speed
    3. features/add-ons/high end analysis
    4. compatability with other packages (c++ etc.)
    5. industry standard
     
  2. sas is popular
     
  3. Not sure if it's popular, worked for them for several years and knew of no customers using it for this (and I searched hard), several banks use it for Risk mgmt (VAR) though.

    On the other way: SAS ETS is very strong at managing time series and I found Base SAS to be unbeatable for backtesting, optimization, walk forward, bootstrapping (edge testing) great degree of control with very few lines of code. The problem is how to learn it: unless you're an academic or you work for a customer/partner getting your hands on it is hard.
     
  4. drtomaso

    drtomaso

    You might consider R- it was originally intended for visualization, but has much of the power of SAS, and is OS.
     
  5. I second that. I do all my backtesting, walk-forward testing, optimization etc. in R, and it works great. You need to get familiar with vectorization though, since loops are very slow compared to e.g. C and Fortran and should be avoided if speed is an issue.

    I think one of the main advantages of SAS is that it does a good job at handling really large datasets that don't fit into memory at once. It seems S-PLUS has some advantages over R in this respect as well (pipelining), although I haven't tried this in practice.
     
  6. nbates

    nbates

    Best language depends on the problem you're trying to solve, the amount of data that needs to be processed and the system's performance requirements.
     
  7. the statistics in matlab is too week.
     
  8. m2n

    m2n

    You can buy the statistics toolbox (or some freeware Matlab library) if you don't find enough statistics functions in Matlab.

    See epchan.blogspot.com for some examples of trading analysis using Matlab.
     
  9. Since crunching any serious measure of numbers is expensive, for runtime efficiency, all of my algorithmic heavy lifting is in C. My gui front-end is in C++ with the Qt framework (Trolltech). Anything else, such as you have mentioned in your poll, would simply not cut it for me.

    -kt