Matlab versus R

tptspecial · Post by **tptspecial** » Wed Sep 05, 2007 5:33 am

CreditCorrelation:

I would suggest that you go with matlab. Not only there is a HUGE library of functions already written, It is USED pretty heavily In hedge funds and investment banks. Citadel for example, uses matlab pretty heavily, especially the high frequency and stat arb groups.

jungle · Post by **jungle** » Wed Sep 05, 2007 2:30 pm

Credit Correlation - have a look at the rMetrics page.

doobs · Post by **doobs** » Wed Sep 05, 2007 7:43 pm

You might want to consider SAS Cry

EDIT: One bad thing about R is the documentation, with Matlab you get excellent documentation and support.

Credit Correlation · Post by **Credit Correlation** » Sun Sep 23, 2007 10:18 pm

Tradenator, urnash, tptspecial, jungle, & doobs

Thank you for your replies.

I asked a friend who has experience with Python, R, and Matlab

(not to mention that he is a C/C++ expert), who suggested using

Python.

Based on his and on other people's advice, I decided to start with

Python, since it is a general purpose language, has a large user

base, public discussion groups, and has a decent number of

packages.

orbital · Post by **orbital** » Mon Sep 24, 2007 4:23 am

Python is a wonderful programming language. So far, I'm still not happy with its plotting capabilities.

R is great for making plots and it's designed to analyze data. The concept of categories and category plots is so useful, I implemented it in the language I'm forced to use, IDL. Point is: play with R's category plot feature.

As a side note, R has bindings to Mondrian which I highly recommend for studying multidimensional data.

urnash · Post by **urnash** » Mon Sep 24, 2007 10:16 am

I do not have any experience whatsoever with Mondrian, but for multidimensinal data I use GGobi, which can be a bit slow, but I do like that you can brush. EDIT: no idea if it is also possible in Mondrian. Oh, looking at the website it seems to be possible...

orbital · Post by **orbital** » Mon Sep 24, 2007 8:58 pm

urnash, i started with ggobi. Brushing is wonderful. However, when I stopped using it, a few months ago, I had gotten annoyed with how buggy it was. (That may have changed in the past 6 months.)

Mondrian just works. It has few features, but the features work :) Bad thing is that Mondrian is extremely strict on its input, which is why you want to use Rserve + Mondrian.

urnash · Post by **urnash** » Tue Sep 25, 2007 11:26 am

[b]orbital[/b]: thanks for your comments. There is a new version of GGobi out, which should be less buggy, but I haven't tested it. I will certainly look at Mondrian in the future, but probably without RServe. This because I've seen that it is not recommended to use RServe on a Windows machine. Can you comment on that? Thanks in advance

orbital · Post by **orbital** » Tue Sep 25, 2007 9:50 pm

Unfortunately, no. I use Rserve on my Linux box, it works great. I'm surprised you use windows, my experience with ggobi on windows was rather painful.

As per Mondrian w/out rserve, here's my one suggestion. Mondrian is fanatical about tab separated files. Read your data into R with read.table, and then write it out with write.csv.

urnash · Post by **urnash** » Thu Sep 18, 2008 2:51 pm

Since this thread has morphed to a thread on R and Graphics, I'll post here the slides of a talk on SpRay, a new R based visual analytics platform that uses both the CPU and the GPU to look at data. Seems very neat. [b]However[/b] I haven't found any other information on it online and I wasn't at the UseR conference, so I do not know more on it. Will keep you posted...