Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Why is Mata much slower than MATLAB at matrix inversion?


From   Nick Cox <[email protected]>
To   "[email protected]" <[email protected]>
Subject   Re: st: Why is Mata much slower than MATLAB at matrix inversion?
Date   Mon, 23 Jul 2012 22:44:58 +0100

I was commenting on Pradipto's comments, which broadened the thread. I completely agree with you that evidence on timings is quite independent of who cites it.

Nick


On 23 Jul 2012, at 22:29, Patrick Roland <[email protected]> wrote:

I'm not quite sure how experience is relevant when comparing something
simple like matrix multiplication across platforms. Presumably
experienced users in Mata and inexperienced users in MATLAB would find
the same discrepancies in runtime.

On Mon, Jul 23, 2012 at 9:57 AM, Nick Cox <[email protected]> wrote:
I am always happy to agree with the idea that one should seek the best in everything.

On axes to grind, or not: That part of the comment was certainly not aimed at you, but just at any one who has axes to grind, including myself. In particular, there is certainly a sense in which I work for Stata, as an Editor of the Stata Journal. If I compared Stata with something else on Statalist, and my comments appeared subjective or biased by my experiences, it would be entirely fair comment to point out that I have Stata roles. This is all part of my main point, that we need to know what experiences lie behind comparisons to evaluate those comparisons.

Nick
[email protected]

Pradipto Banerjee

Nick, that's true. But, if you allow me to add (as a Stata "beginner"): both Stata and MATLAB can be easily integrated to achieve the best of both, e.g. in Stata -winexec- or -shell- and vice versa in Matlab -system- command, and I don't see why not use the best of both ... (really there is no "axe to grind" - I don't work for either StataCorp or Mathworks).


-----Original Message-----
From: [email protected] [mailto:owner- [email protected]] On Behalf Of Nick Cox
Sent: Monday, July 23, 2012 11:29 AM
To: '[email protected]'
Subject: RE: st: Why is Mata much slower than MATLAB at matrix inversion?

Broad commentary on different software is kind of fun but I think some ground rules are needed if it is also to be considered serious.

That is, what backgrounds do people bring to this discussion? I get the impression that Pradipto is a beginning Stata programmer and an experienced MATLAB (that's still its name) user-programmer and his remarks are to be interpreted accordingly. It would be easiest if there were people who have spent approximately similar time programming MATLAB and Stata, were equally competent in both and had no axe to grind. Those people are likely to be thin on the ground; that being so, comments are difficult to interpret without some idea of people's backgrounds.

Nick
[email protected]

Pradipto Banerjee

I had the same issue. I think different applications have their pros and cons. Both Stata and Matlab have their places.

Stata is great for data manipulation and data visualization, merging databases, or trying to quickly see whether a few variables are related to others, carrying out variety of regressions both across time and cross-section, i.e. primarily to build insights from a database without first building a whole set of tools around a database.

On the other hand, once all the insights are developed in Stata, Matlab is perhaps preferable to build the rest of the application because it is faster, has many use toolboxes like optimization, integrates well with the engineering & financial world (e.g. Bloomberg, lots of financial databases & APIs), programming is much easier, awesome editor and is very good as a single environment to develop a complete package.

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index