What Software Is Best for Fitting Data Distributions?

Click For Summary

Discussion Overview

The discussion centers around the best software options for fitting data distributions, exploring various tools and methodologies used in statistical analysis. Participants share their experiences and preferences regarding software transparency, usability, and suitability for different tasks.

Discussion Character

  • Debate/contested
  • Technical explanation
  • Exploratory

Main Points Raised

  • One participant emphasizes the importance of using software that allows access to source code, arguing that software that conceals its processes has limited scientific utility.
  • Another participant counters this view, suggesting that many widely used software packages (e.g., Mathematica, MATLAB) do not provide full source code yet are trusted by users.
  • A suggestion is made to explore statistical packages that handle common distributions, with an alternative approach of using a chi-square goodness of fit test for unsupported distributions.
  • A participant shares their experience using MATLAB and GNU/Octave, noting the importance of transparency and the ability to replicate results, especially in published science.
  • Concerns are raised about potential pitfalls in software handling of data, with a recommendation that the choice of software should depend on the significance of the analysis and the scrutiny it may face.
  • Some participants express a preference for custom-written software or scripts tailored to specific tasks for maximum transparency and control.

Areas of Agreement / Disagreement

Participants express differing views on the necessity of source code access for software used in data fitting. While some advocate for transparency and custom solutions, others argue that established software can be reliable despite lacking full source code. The discussion remains unresolved with multiple competing perspectives.

Contextual Notes

Participants highlight the importance of replicability in scientific research and the varying degrees of scrutiny that different analyses may require. There is an acknowledgment of the limitations of certain software in handling specific data fitting tasks.

Mark J.
Messages
81
Reaction score
0
Hi.
Which is the best software to use working on data fit distributions from your experience?

Regards
 
Mathematics news on Phys.org
Anything I write myself, followed by anything I can get the source code for.
Software that hides what it is doing has limited use in science.
 
Simon Bridge said:
Anything I write myself, followed by anything I can get the source code for.
Software that hides what it is doing has limited use in science.

You can't be serious.

People use software packages all the time that do routines that we trust where we do not have the full source code. You think we have the source code for Mathematica, Maple, SPSS, SAS, MATLAB, JMP and so on?

For the OP look at some statistical packages which may do it for common distributions.

If your distribution is not supported one idea would be to use a chi-square goodness of fit test.
 
When I last used matlab, you could look in the m-files but not at the engine.

For my thesis I would set up the simulations in MATLAB or gnu/octave and when I thought I had something I would convert the whole thing to C.

I had colleagues who were always getting caught out by some little wrinkle that MATLAB or whatever did not handle quite right.

What you use depends on what you want to do - how important is it and how much scrutiny will it be subject to? But the very best, which was the question, is going to be specifically tailored to the task and completely transparent .. eg. something you wrote yourself (or hired someone better than you to write for you.)

You don't need the source code if you don't expect to get any unusual results. But if someone is likely to ask if what you found is really there or just an artifact of the computer program - you need the source. For publishes science you need, on hand, everything another researcher needs to duplicate your results.

These days gnu/octave fits my needs ... especially with the power of modern computers. The skills using it translate to other scripting languages. I know people who swear by spreadsheets ... which I personally find painful.
 

Similar threads

  • · Replies 14 ·
Replies
14
Views
1K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 6 ·
Replies
6
Views
1K
  • · Replies 2 ·
Replies
2
Views
3K
Replies
6
Views
1K
  • · Replies 2 ·
Replies
2
Views
1K
  • · Replies 1 ·
Replies
1
Views
1K
  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 11 ·
Replies
11
Views
3K
  • · Replies 2 ·
Replies
2
Views
1K