I have always used MATLAB for modeling; however, many of my affiliates cannot afford buying 2.1K license for it and spending 1K for each of the additional libraries. Someone recently suggested to me SCILAB that is apparently widely used in Europe. Are there any users of SCILAB on this forum? Any positive/negative experience to share? I would like to learn more about this product before I suggest it to my colleagues. It looks like one could even run some of the MATLAB scripts on it. Please share your experience. Cheers, MAESTRO
Matlab is a far better, far more reliable tool. I can't imagine anyone choosing Scilab as a replacement for anything critical. $2k for a license is, for a professional user, cheap.
You might also consider R, which is widely used in Financial modeling. For some good examples, see http://blog.fosstrading.com/ I think you are far better off with R than with a clone of Matlab, and in fact probably better off than with Matlab itself.
I would also vote for R. Rserve (TCP/IP server) provides an interface to R and there are clients for C/C++, PHP and Java. I use the Java client and think it works fine. I imagine the Java client can be used in a seamless fashion with Scala, Jython et al. Henrik
Matlab (core) will go up to $2.2k US in January. MathWorks gets you with the core and all the linked toolboxes. For the computational finance case, that adds up to an additional $7.6k US plus about 18% of the total price for the annual software maintenance service. It is too bad MathWorks does not have a personal use program similar to Mathematica ($250 this holiday season) SCILAB (and it's forked competitor Scicoslab) is great for me since it includes something I really want that is similar to Matlab's SIMULINK (another $3.15k) called xCOS. And it is Open Source and essentially GPL, more appropriately Free Libre Open Source Software (FLOSS). Some m-files are compatible, and I think Scilab has a series of tools that can read/convert the m-files. Unfortunately, if your m-files uses the MathWorks toolboxes, I doubt it is fully compatible. Since I use Matlab/Simulink at work, I can say that Scilab is just not fully compatible. There are some things in Scilab that are done better than in Matlab, I recall an m-body celestial orbital model for interplanetary travel that is more efficient in Scilab that was developed by the French. I also demonstrated in Rev2011a a non-vectorized routine using for-next loops that was faster than the vectorized version in Matlab, but vectorized version in Scilab was faster. Again, that may just be a simple quirk or a luddite programmer error that I stumbled on, and it may be fixed in Rev2011b (Matlab released this last September as part of the sotware maintenance service). If you can read French or German, there are plenty of references to get you up to speed. Alas, I am not in that crowd so that is the major drawback for me. There is only one book that barely covers Scilab for beginners in English (Modeling and Simulation in Scilab/Scicos with ScicosLab 4.4). I think if Scilab matures and gain critical mass in areas that primarily speak English, it will give Matlab a run for their money, in my humble opinion. I've heard the Asian continent is starting to be aware of Scilab. However in the interim for English only readers, I've attached an 8+ MB pdf that summarizes in detail Scilab that is in English for your perusal. Update: Looks like EliteTrader does not like 8+ Meg files, so here is the PowerPoint link to the file: http://www.heikell.fi/downloads/scilab.ppt Hope this helps, dsss27
As far as I am aware the following matlab code does not have an equivalent in scilab: [s stat] = urlread(['http://finance.yahoo.com/q?s=SLB']); I use matlab R2008b and the function 'urlread' is what it makes it useful for obtaining EOD data and automating scans. Without an equivalent function scilab is useless to me, but I expect that Excel VBA would have something similar, although I have not looked into it.
Not to dissuade anyone from using any particular one of these excellent environments, but just by way of a point of comparison, there are extensive capabilities in R's quantmod package for downloading EOD data. getSymbols("GOOG",src="yahoo") getSymbols("YHOO",src="google") http://www.quantmod.com/ There is another package for doing some Interactive Brokers automation.
Thank you. I agree with that. However, I wasn't trying to use it for trading or even modeling of the real market data; I was trying to use it as a teaching tool for my students. I am trying to lead a group of beginners to a deeper understanding of randomness and its unusual properties. I have developed a few models that are based on the correlated, biased, and abnormally distributed random walks that I would like my students to research. Most of them, however, could not afford MATLAB and I am strongly against pirating! So, from this point of view do you think SCILAB is still not good? Cheers, MAESTRO