Serving the Quantitative Finance Community

 
User avatar
VegaPlus
Topic Author
Posts: 0
Joined: November 7th, 2008, 9:45 pm

Should I learn Scilab or R?

January 5th, 2011, 3:20 pm

Hi,I am a graduate and would like to prepare for future interviews in quantitative fields (front office quant, model quant, stat arb trading, ...). I don't know Matlab so I'm looking for a free alternative solution. I know that Scilab is very similar to Matlab but there exist some differences in particular functions. So I don't know if "Scilab" on a CV would be seen as "Matlab"... But I can see on job offers that many HF/banks are asking for R knowledge. As R is free as well, I'm wondering if it is no better to learn R than Scilab.I don't have a particular need right now, it's really for upcoming interviews. But I have a good knowledge of Java or C++.And let's say I go for R, what kinf of questions can they ask me to check my R knowledge so that I can go straight to the key points intead of wasting time on useless features.Thanks.
 
User avatar
mxyzptlk
Posts: 0
Joined: August 25th, 2009, 4:33 pm

Should I learn Scilab or R?

January 5th, 2011, 6:59 pm

Without a doubt learn R. You may also consider Octave as an open-source alternative to Matlab with very similar syntax.Since you mention many types of quant positions, it is hard to say what the proper subset of R is. Take a look at the CRAN Task View: Empirical Finance for an idea of relevant packages. You may also want to learn about integrating R & C++ and possibly R & Excel. There are some good books on R such as general (The R Book by Crawley, R in a Nutshell by Adler) and many specific ones (like Springer series which has time series, bayesian, monte carlo, ...).
 
User avatar
Hansi
Posts: 41
Joined: January 25th, 2010, 11:47 am

Should I learn Scilab or R?

January 5th, 2011, 7:13 pm

I think R might be the best option. Most people are familiar with Matlab and/or R but very few people I know have even touched Scilab or Octave.R is becoming more and more popular from what I've heard for prototyping, scripting and quick analytics's since it's open source and quite easy to work with.I've never liked Matlab's black box binaries for a lot of the toolbox and others agree with me that this is an issue.I had an interview for a quant position where R was a requirement recently. The questions were pretty basic just to make sure I'd used it. Some syntax questions, stat functions, integrating with Excel etc. Getting a R book might help but I think grabbing a copy of the files here and having a look at them is a good place to start off with R: http://stats.lse.ac.uk/penzer/ST419materials/ and going through this too: http://www.cyclismo.org/tutorial/R/If you know about the stuff in the tutorial, the stuff in the class notes, how to use RExcel and maybe call R from C++ and C++ from R, I think you should be fine for pretty much anything that might come up in an interview.Also grab a copy of Tinn-R, somewhat better than using a stand alone text editor in some cases. Notepad++ and the R syntax and autocomplete plugin is good too.
 
User avatar
frattyquant
Posts: 0
Joined: March 4th, 2010, 8:10 am

Should I learn Scilab or R?

January 6th, 2011, 11:22 pm

I'd say learn Octave. The syntax is almost exactly the same as Matlab, which is by far the most popular "modelling environment" in banks. R has a lot of great features and ready made code, but working with its data structures can be a real headache.
 
User avatar
Hansi
Posts: 41
Joined: January 25th, 2010, 11:47 am

Should I learn Scilab or R?

January 7th, 2011, 5:21 pm

The problem with Octave is that it doesn't have Matlab's toolboxes and that's mainly what they want you to know about when it comes to using Matlab in banks and funds from what I've heard. If you want to have Matlab on your CV then going straight for Matlab is best.
 
User avatar
VegaPlus
Topic Author
Posts: 0
Joined: November 7th, 2008, 9:45 pm

Should I learn Scilab or R?

January 9th, 2011, 12:28 pm

Thanks for your answers. I think I'll go for R now because it's massively asked in quant roles. It looks pretty easy, though it has a disgusting syntax!
 
User avatar
Costeanu
Posts: 1
Joined: December 29th, 2008, 5:33 pm

Should I learn Scilab or R?

January 12th, 2011, 9:58 pm

VegaPlus, When I was a grad student I had access to cheap software through the university. I don't remember exactly what I bought, Matlab or Maple, but I think it was around $100. If you can find such a deal, I think Matlab should be a good investment. In banks people rarely use Matlab, they use C++. However, with Matlab you can explore more quickly numerical stuff. While you are still preparing for future interviews, you need to code things with your own hands, like a Monte Carlo scheme, a tree, a finite difference, see what problems you encounter when you price barrier options, or when you build a tree for a local vol model. With C++ you'll be able to do 3 such projects, with Matlab 10, let's say. Much better to have 10 than 3. Also, R is a good investment of time if you want to learn statistics. Just be aware that quants on the sell side don't use statistics that much. If you want to use R for numerical projects, forget about it, it's too slow (at least in my experience). I used SciLab in the past. It's ok, if you really need to deliver something, and have no other tools. But it really can't compare to Matlab in terms of documentation. And when you want to learn something you do care about the documentation. In terms of performance, scilab is quite good; I needed to use some large sparse matrices and it did the job well. So, if you know Matlab, go and work for a corporation where you don't have access to it, and use Scilab for a quick job, you will do ok. Learning Scilab for itself doesn't make sense in my opinion. Best,V.