[R-SIG-Finance] Matlab vs. R performance/convenience benchmarking for quantitative analysis business

[Tobias Verbeke]

Hi Samuel,

We have been integrating R in corporate environments
(finance, pharmaceuticals, media, ...) for many years
and have done much more migrations from Matlab to R than
the other way around.

Some of the advantages:

1) the statistical community is using R as its lingua franca;
you therefore have much more chance to take benefit of novel
methods as R packages appear with the papers and as these are
likely to be taken up only later by the Matlab world.

2) you avoid all hassles of license management and have no
license fees to pay for server use of R whatever number of
CPUs, users, etc.

3) R packages have a much richer structure than Matlab toolboxes
and out-of-the-box support inclusion of technical documentation, 
functional user guides, automated software testing etc. inside
one standardized unit.

4) if there is a need to interact with suppliers or research
partners, R does not imply any threshold for these partners as
it can just be downloaded from the internet.

Matlab does have the advantage that you can make executables
of Matlab code to distribute to end users, but in R world this
type of problems is more commonly solved by working in
client-server mode (with desktop clients or web clients)
or embedding R inside the application.

In terms of the language, R is much more pragmatic than Matlab
and its data types are more suited to statistical analysis
(categorical variables, ordered categorical variables) than
the Matlab ones (where a plain string e.g. is represented as
a matrix). In some other domains of applied mathematics,
though, (say computational fluid dynamics or engineering)
Matlab is likely to have more functionality to offer.

With respect to the benchmark, it seems to be quite outdated
and besides progress made in recent years, there is quite a
number of tools to tackle performance issues if the application
requires it

http://cran.r-project.org/web/views/HighPerformanceComputing.html

Also, R can be used to conveniently wrap C, c++, Fortran or Java
code for parts of code where performance is important.

In terms of IDE, the Eclipse/StatET environment is the equivalent
of the Matlab GUI and used by most of our customers.

Best,
Tobias

-- 

Tobias Verbeke
Manager

OpenAnalytics BVBA
Kortestraat 30A
2220 Heist-op-den-Berg
Belgium

E tobias.verbeke at openanalytics.eu
M +32 499 36 33 15

http://www.openanalytics.eu