4 ms·

JavaScript library for the visualization of statistical distributions

49 points by r0muald 11y ago

lorenzhs 11y ago
Very nice! One thing that would make it easier to see how the distribution behaves would be to have a number of common parameter choices as presets for each distribution, like the Wikipedia pages often do. At least, that is what I would find useful when picking a distribution to use for random inputs. It's also what bugs me about every documentation of C++11's <random> header I've found---they're great if you know which distribution you need, but picking one (when you have the freedom to choose) is hard. Of course, that isn't the job of a language reference, and this could be the tool for that.
stared 11y ago
The actual interesting link is here: http://learnbayes.org/demo/stat-distributions-js/distributionTable.html http://learnbayes.org/demo/stat-distributions-js/distributio... (a table of distributions).
javajosh 11y ago
If I could send two things back in time to aid my undergrad physics self in his studies, it would be this and http://acko.net/blog/how-to-fold-a-julia-fractal/ http://acko.net/blog/how-to-fold-a-julia-fractal/ (which is as close to an intuition about complex numbers as I've every seen written to a screen). Really useful and cool. Thanks.
graycat 11y ago
Please: (1) Yes, in probability, distributions are crucial concepts. (2) In practice, for random variables taking real number values, usually can't know the distribution. For random variables taking vector values of dimension over, say, three, essentially never can know the distribution. (3) The main approach in practice is to use the classic limit theorems, especially the law of large numbers and the central limit theorem, to let us know at most only a little about the distributions. (4) More generally, we have to work nearly always knowing few or no details about the distributions. Looking at some histograms? For a first-cut, sometimes maybe okay. Looking at pictures of distributions? Nearly useless.
- stared 11y ago
  For Bayesian modeling making assumptions about distributions is crucial.

49 points by r0muald 11y ago

Very nice! One thing that would make it easier to see how the distribution behaves would be to have a number of common parameter choices as presets for each distribution, like the Wikipedia pages often do. At least, that is what I would find useful when picking a distribution to use for random inputs. It's also what bugs me about every documentation of C++11's <random> header I've found---they're great if you know which distribution you need, but picking one (when you have the freedom to choose) is hard. Of course, that isn't the job of a language reference, and this could be the tool for that.

stared 11y ago

The actual interesting link is here: http://learnbayes.org/demo/stat-distributions-js/distributionTable.html http://learnbayes.org/demo/stat-distributions-js/distributio... (a table of distributions).

javajosh 11y ago

If I could send two things back in time to aid my undergrad physics self in his studies, it would be this and http://acko.net/blog/how-to-fold-a-julia-fractal/ http://acko.net/blog/how-to-fold-a-julia-fractal/ (which is as close to an intuition about complex numbers as I've every seen written to a screen). Really useful and cool. Thanks.

graycat 11y ago

Please: (1) Yes, in probability, distributions are crucial concepts. (2) In practice, for random variables taking real number values, usually can't know the distribution. For random variables taking vector values of dimension over, say, three, essentially never can know the distribution. (3) The main approach in practice is to use the classic limit theorems, especially the law of large numbers and the central limit theorem, to let us know at most only a little about the distributions. (4) More generally, we have to work nearly always knowing few or no details about the distributions. Looking at some histograms? For a first-cut, sometimes maybe okay. Looking at pictures of distributions? Nearly useless.

stared 11y ago
For Bayesian modeling making assumptions about distributions is crucial.