5

I have only a little experience in math calculation tools. So please don't be strict if my question is not well defined :)

Which tool (Wolfram Mathematica, MATLAB or something else) is better to use for statistics? Where can I find quickstart guides and tutorials for it?

Mihran Hovsepyan
  • 167
  • 1
  • 4
  • I'm voting to close, not because it's a "bad question"...but I because it is likely that it will solicit an opinionated debate about the merits of different software. People become loyal, attached to various options. –  Nov 07 '12 at 17:28
  • 5
    To prevent this question from being closed as potentially argumentative or too vague, please edit it to indicate *what* applications you contemplate and what performance criteria you have. "Statistics" is far too broad to admit a coherent or useful set of replies. – whuber Nov 07 '12 at 18:53
  • -1 What you are stating is a problem of choice and thus a decision to be made. Doing so without providing preferences and goals to measure value of alternatives is unfortunately rather pointless or degrades to a simple matter of taste. Too bad. – gwr Nov 14 '15 at 13:01
  • The correct answer is R. Don't waste your money on proprietary tools unless you are sure you need them (which is almost never). – Marc Claesen Dec 12 '15 at 08:51

4 Answers4

10

Wolfram Mathematica is a very capable software for doing statistics, and unlike Matlab, its statistical functionality is included in the core Mathematica.

Sasha
  • 407
  • 4
  • 13
  • 4
    I'd like to see a reference for this assertion: "Mathematica is often faster than R for statistical data crunching." – Zach Nov 07 '12 at 19:10
  • 2
    Or this one: "Mathematica has superior statistical visualization functionality". Nothing at the linked site convinces me of that. (None of which is to imply that Mathematica isn't worthwhile.) – gung - Reinstate Monica Nov 07 '12 at 19:16
  • 5
    And let's not forget, if you (the OP) will start using it, there's a [StackExchange for Mathematica](http://mathematica.stackexchange.com/) – Szabolcs Nov 07 '12 at 19:42
  • 3
    I guess to make it substantial we should make a comparison across the board, but my claim was based on `AbsoluteTiming[a=RandomVariate[NormalDistribution[],10^7];]` reports 072 seconds, while `system.time(a – Sasha Nov 07 '12 at 19:51
  • @Zach In addition to the comment above, here are some timing for the linear algebra. `a – Sasha Nov 07 '12 at 20:08
  • I agree, Mathematica is faster than R. Not because Mathematica is very fast but because R is very slow, and it's not very well organized. But it's free. – skan Apr 02 '13 at 00:47
  • Superior graphics ? http://gallery.r-enthusiasts.com/ – Stéphane Laurent Apr 02 '13 at 06:44
  • @Zach: I think Sasha gave more than sufficient proof for his assertions now. Do you have a counter example to show R is as fast or faster? – curious_cat Apr 02 '13 at 07:05
  • 1
    I think if the matter of speed in Lin.Algebra comes into play, Matlab is the fastest of the 3, at least on average. Also given that R through the `Matrix` package has access to `Csparse` and `CHOLMOD` and similar is in a worst case scenario as fast as anything else when it comes to sparse matrices. (And clearly you have access to packages like `RcppEigen` that are help with small dense matrices also). Despite its merits on symbolic computation *Mathematica* is _not_ the lingua franca of Statistics, *R* is. – usεr11852 Apr 02 '13 at 13:24
  • @usεr11852: Now, it could very likely be argued that the *lingua franca* of statistics - and even more generally of solving problems by mathematical model is the international language of mathematics. So one criterium of choice could then be: which tool stays closesr to mathematical notation and procedures? CAS like Mathematica and Maple certainly have a very big advantage here especially since this capability niwadays does not mean that you have to forfeit on performance! – gwr Nov 14 '15 at 12:58
  • @gwr: When referring to *lingua franca* I am concerned with computing environments here. For better or worse R is the software used more frequently in Statistics. Its depth and breadth is unparalleled. Just take a look at: https://cran.r-project.org/web/packages/available_packages_by_name.html and compare that list with the one of the 3 M's computing suites have to offer. Quality-wise R packages are on average *almost surely* worse than their M counterparts (professional devs are *almost surely* better devs than non-professional ones) but we are not examining this aspect for a *lingua franca*. – usεr11852 Nov 14 '15 at 20:21
  • Let's not forget that ther is [RLink](http://reference.wolfram.com/language/RLink/guide/RLink.html) for *Mathematica* so all of those packages for R can be used rather easily. If your world and the problems you solve are broader than statistics, the nothing beats a package like Mathematica. Also time is limited: Why learn another language if there is some kind of *lingua franca* for mathematical problems out there? – gwr Dec 12 '15 at 08:07
9

For statistics I would recommend R - computation environment concerning statistics.

Here you can find many tutorials for R.

qmsource
  • 191
  • 1
  • @Mihran R and MATLAB have a lot in common, so if you have any experience with MATLAB, I think you would do fine with R, too. Thanks for the link qmsoure, I've been looking for something like this for a few days. –  Nov 07 '12 at 17:24
2

There are specific tools out there for statistical analysis. For example, R, SPSS, and Minitab.

Nevertheless, both Mathematica and MATLAB are capable of doing the required computations; which is "better" is a matter of taste and application requirements, most of the time.

MATLAB doesn't ship default with many statistical subroutines. However, it does have a Statistics toolbox that contains much of the functionality found default in some other packages. It all depends on what exactly you're trying to do, statistics wise, and what your comfort level with programming is. Maybe you can elaborate more on what you're attempting.

(Of course, the man-hours spent re-programming an algorithm for, say, ANOVA is often costlier than just buying the license to the toolbox that has a stock routine outright).

Finally, to answer your question, I personally prefer MATLAB. I find that it is easier to program in and allows for easier rapid development and testing of algorithms, etc. Syntactically, I prefer MATLAB. MATLAB likes to keep things close to pseudo-code. Mathematica tries to keep things closer to mathematical notation.

You can find MATLAB tutorials all over the internet. Just google "MATLAB tutorial" and you'll have a slew of results.

Emily
  • 152
  • 6
1

Some people would say that Matlab is for numerical matrix calculations and simulations, and Mathematica for symbolic calculation. But nowadays both can perform both things. And we should add another contender, Maple.

In my opinion Matlab is too big and needs too much resources. You should only choose it if your company or school forces you or if you need some rare simulation tool. If not, you should choose Mathematica or Maple. They are also very good with graphics, numeric matrixes and both have simulation tools. But they have something much better than Matlab: symbolic and abstract mathematics.

If you can't afford commercial software or you like free software you can try R, for statistics and little bit more.

gwr
  • 256
  • 5
  • 16
skan
  • 814
  • 1
  • 7
  • 20