Template talk:Probability distributions

Not Italic

This template is a good idea, but I think we should not have italics in such a small font, its hard to read. See for example Template:popgen. PAR 00:27, 24 April 2006 (UTC)

Good call, I removed them. --Btyner 01:48, 24 April 2006 (UTC)

"degree" distribution?

Is this a true distribution? It seems not to be, or certainly not descibed as one. Baccyak4H (talk) 02:36, 19 November 2006 (UTC)

No, judging from the definition of degree distribution. Removed. --MarkSweep (call me collect) 03:41, 19 November 2006 (UTC)

And again, is the "fading" distribution in the same boat? Baccyak4H (Yak!) 20:08, 12 December 2006 (UTC)

I had the same thought (before looking at this talk page), so i'll go ahead and remove it. Qwfp (talk) 09:42, 30 January 2008 (UTC)

Benford's law

I added a link to Benford's law, which itself implies a very well defined discrete distribution. I do note that that page is not of similar format to most other distribution pages, so if some think its inclusion in this template is problematic, let's discuss. Baccyak4H (Yak!) 19:49, 27 February 2007 (UTC)

missing

frechet and log-gamma are missing... --84.75.21.114 12:41, 31 March 2007 (UTC)

spaces in wikitext

I boldly changed the spaces in the displayed names of the distributions to non breaking spaces. This way, a distribution's name will always appear on one line. The only drawback I see is that the line breaks may become slightly more ragged.

If this displays poorly on same pages or in some contexts, it can be reverted, and apologies in advance. Baccyak4H (Yak!) 18:46, 31 October 2007 (UTC)

Hotelling's T-square

The Hottelling's T square is a multivariate generalisation of the Student's t distribution. Therefore, I think it should be list in the column of the multivariate distributions in this template. 08 Dec 2007 —Preceding unsigned comment added by 129.111.194.230 (talk) 02:37, 10 December 2007 (UTC)

Although it's used in multivariate statistics, Hotelling's T-statistic is a scalar so Hotelling's T-distribution is univariate so it's correct where it is. --Qwfp (talk) 16:19, 24 February 2008 (UTC)

Too large

While i think this navbox is useful, i'd suggest it has grown too large, which limits its navigational value -- see WP:NAVBOX#Properties. In particular the number of univariate continuous distributions is a bit overwhelming. Seems worth discussing which of the three solutions suggested at WP:NAVBOX#Properties might be appropriate. Personally i'd favour the last, "Link only the immediately related articles while hiding the rest, like in the case of {{Johnny Cash}}." This would require first creating more subdivisions, at least for the univariate continuous distributions. Could use the same ones used by Probability_distribution#List_of_important_probability_distributions i.e. categorise by support.

The other option that suggests is to limit the navbox to "important" distributions, but I don't like that option as it would open the door to endless discussion of what constitutes "important".

I also think the "Miscellaneous" grouping needs some cleaning up and pruning:

• A "Families" grouping could contain location-scale family, exponential family, Tweedie distributions, Pearson distribution (...?).
• Other entries such as prior distribution, sampling distribution, bimodal distribution are uses or descriptions of distributions rather than being distributions themselves and I don't feel really belong in the navbox at all.
• In principle for consistency maybe should have a "Singular" grouping to contain Cantor distribution.

Having put this all down in writing it looks like quite a lot of work though, especially as the "{{Johnny Cash}}" solution would require editing each of the pages that transclude this navbox. I don't have time to do it at the moment and whether i ever attempt it probably depends on the reaction of others here. Qwfp (talk) 11:33, 30 January 2008 (UTC)

Yeah, it has grown a bit unwieldy. Please be bold! Btyner (talk) 18:20, 2 February 2008 (UTC)

I concur the whole thing is too big. I may have a go soon; I like the idea of collapsable boxes so will probably try something along those lines. If I do try, I'll start in my userspace and I'll link it here. Baccyak4H (Yak!) 21:06, 4 February 2008 (UTC)

Please, go ahead. I don't think i'm ever likely to get around to it myself to be perfectly honest. I'll watch with interest. Qwfp (talk) 22:12, 4 February 2008 (UTC)

I got around to it after all, inspired by the discussion over at WT:WPM#The numbers template. The result is at User:Qwfp/tempprobdist for now. An example:

{{User:Qwfp/tempprobdist|multivariate}} 

produces: User:Qwfp/tempprobdist

Let me know what you think. I'll also run it past WT:WikiProject Probability (if in case anyone's still watching that) and WT:WikiProject Mathematics before replacing {{ProbDistributions}}. Qwfp (talk) 15:24, 24 February 2008 (UTC)

 Done. I've replaced it with the new and somewhat more compact version with collapsible groups. Qwfp (talk) 11:48, 1 March 2008 (UTC)

Too non-layperson

I may not have a BS in statistics, but it was pretty hard to figure out where something like normal distribution would be found in 'Continuous univariate supported on the whole real line'. If the navbox is to facilitate navigation, should its purpose be somewhat to help the uninitiated find information? If it's going to include every topic under the statistics sun, then the headings can have a huge impact on the information flow, so to speak. Rhetth (talk) 22:46, 4 August 2008 (UTC)

I agree, there is space for a couple of examples, we should give them. --BozMo talk 20:33, 3 February 2009 (UTC)

I'm not entirely sure what either of you are suggesting the headings should be. The current headings are based on those in list of probability distributions. Whether there's space for longer headings in a single line depends on the width of your browser window – some of the headings are quite long already. Something else which might help the 'uninitiated' would be to have a simpler, smaller navbox containing only a couple of dozen of the more commonly-used and commonly-viewed distributions, such as currently in my sandbox. This could be used on the pages for these distributions. Comments? Qwfp (talk) 21:16, 3 February 2009 (UTC)

I've since created such a smaller simpler navbox as Template:Common univariate probability distributions Qwfp (talk) 13:37, 1 July 2009 (UTC)

The Dirac Comb is a probability distribution

The Dirac comb of period 2 pi is to circular statistics what the Dirac delta function is to linear statistics. If the Dirac comb is to be eliminated as a probability distribution, then the Dirac delta function must be eliminated also. The Dirac comb of period 2 pi is the limiting case of the wrapped normal distribution, the wrapped Cauchy distribution, the wrapped Levy distribution, all of which are circular PDF's. PAR (talk) 21:20, 27 March 2010 (UTC)

If the Dirac comb is a probability distribution, then that article could describe its properties as a probability distribution. But it does not. I would be happy to remove Dirac delta function from this template as well, and say anything that needs to be said at degenerate distribution (which could easily be extended to also cover limits from continuous distributions, with suitable links from other articles as needed). By the way, is a Dirac comb of period 1 a probability distribution? --Rumping (talk) 19:47, 28 March 2010 (UTC)

In the last half year, nothing has happened to describe the properties of the Dirac comb as a probability distribution.--Rumping (talk) 09:38, 23 October 2010 (UTC)

Back on directional statistics: The Dirac comb of period 2π is to directional statistics as the Dirac delta function is to linear statistics. It is a "wrapped" Delta function. I think the Dirac Delta should have a place as a probability distribution, as well as the 2π Dirac Comb, but I'm not sure how to do it properly. I think the Dirac Delta is presently miscategorized (I don't think it degenerate which applies to discrete only.) Can a function be in two categories? Dirac Delta and 2π Dirac Comb would both be in something like "univariate, continuous, support of measure zero" but 2π Dirac Comb would also be in "Directional" as well. Any suggestions? PAR (talk) 06:59, 31 December 2015 (UTC)

Fermi Dirac and Bose Einstein are not probability distributions

Well, it should be pretty obvious from their lack of unity normalization, but the Fermi-Dirac distribution function and Bose-Einstein distribution function are not probability distributions. Rather they represent an expectation value of the number particles in a given energy level.

• The actual distribution for the number of particles in a Fermi-Dirac level is a simple Bernoulli distribution -- only 0 or 1 particles allowed in a level.
• The actual distribution for the number of particles in a Bose-Einstein level is a simple geometric distribution. Probability(N) = const * probability(N-1).

This is just a notice and explanation, I'm going to make the edits. --Nanite (talk) 17:04, 25 May 2013 (UTC)

Semi-protected edit request on 21 May 2018

Please change the "Probability distributions" table at the bottom of the page to include a link to the "Multivariate generalized beta distribution" subsection of the "Generalized beta distribution" page (https://en.wikipedia.org/wiki/Generalized_beta_distribution), in the "Multivariate (joint)" cell as part of the "Continuous" list in that cell.

Please also add a link to the "Generalized beta distribution" in the "Families" cell of the same table. 128.187.112.9 (talk) 16:28, 21 May 2018 (UTC)

 Not done: please establish a consensus for this alteration before using the {{edit semi-protected}} template. Eggishorn (talk) (contrib) 18:39, 22 May 2018 (UTC)

Semi-protected edit request on 20 April 2023

Change “Erlangl” to “Erlang” (typo). 134.95.142.19 (talk) 08:44, 20 April 2023 (UTC)

 Done M.Bitton (talk) 13:29, 20 April 2023 (UTC)