Wikipedia talk:WikiProject Statistics

WikiProject Statistics
Main page	Talk page	Members	Templates	Resources

Consistent notation of the quantile function

I noticed that different notations are used to denote the quantile function of probability distributions. The most common ones are $Q(p)$ (e.g. Tukey lambda distribution), $Q(u)$ (e.g. Dagum distribution), $F^{-1}(\cdot )$ (e.g. Kumaraswamy distribution), and $x(F)$ Laplace distribution. In other cases, the quantile function is redubbed as "random variate generation", and described as a transformation of a standard uniform random variable (rv), i.e. $U\mapsto X$ (e.g. Burr Type XII distribution).

Although I realize that this is similarly the case in the literature, I believe that arbitrarily inconsistent notation like this, can be very confusing to readers, especially newcomers. But unlike the published literature, here, the notation can be made consistent.

I don't have a strong preference myself, but I (usually) prefer $x(F)$ over $F^{-1}(\cdot )$ nowadays, since both $Q(\cdot )$ and $F^{-1}(\cdot )$ have no standard for the name of the parameter (I've seen $p$, $q$, $u$ and $y$ used in different places). I can also see that standard uniform transformation notation, e.g. $X=-\log(1-U)$ for the standard exponential distribution, could be a good choice, but only if it is described as being the "quantile function", and not only "random variate generation". jorenham (talk) 14:26, 14 December 2023 (UTC)

If we were going to try to standardise, I would prefer Q(u), particularly because it leads naturally to q(u) for the density quantile function. But note also that adopting a standard on Wikipedia would make it inconsistent with the way the quantile function is usually presented in the literature for some distributions. Newystats (talk) 02:34, 15 December 2023 (UTC)

Coupling the QF notation with that of the QDF makes sense to me.

I guess that notational inconsistency with the literature is inevitable. I even stumbled accross a $G(p)$ today (C.L. Mallows, '73).

Perhaps it's a good idea to explicitly list the common QF notations on the quantile function page? jorenham (talk) 02:54, 15 December 2023 (UTC)

Should R not be listed as of top importance?

I see some statistics software (e.g. Minitab) has an importance of mid-importance, but there's nothing for R, which I don't think reflects the importance R does have. Drkirkby (talk) 15:43, 3 February 2024 (UTC)

Notability of John H. Wolfe

The article John H. Wolfe has gone through a PROD, but still has issues as it is based on one secondary textbook claim that his work on model-based clustering matters. It was created directly by a novice editor (Stat3472 33 edits). The article model-based clustering supports him as the inventor, but whether this is big enough for notability is unclear. Comments on the talk page please, perhaps better than AfD. Ldm1954 (talk) 09:56, 2 March 2024 (UTC)

094746 102.213.69.202 (talk) 13:42, 20 March 2024 (UTC)

Conjugate priors: raise importance to Top or at least Mid? And needs expert attention

I am not a statistician but I need to understand conjugate priors for my research. Depending on what the scope of ``its field" is meant to be in the Wikipedia guidelines for importance rating, it seems to me this should be anything from a Top to a Mid priority article, but not a Low priority article as currently rated. Here is a quote from a wikipedia page that characterizes

At Wikipedia:Version 1.0 Editorial Team/Release Version Criteria

Top priority is described as: Subject is a must-have for a print encyclopedia

I think that conjugate priors fall into this category. (I find the priority critera on this page too vague, however.) Anybody who reads the literature on the theory or the application of Bayesian statistics is likely to run into the term fairly quickly.

Another page, which I have closed and haven't been able to easily track down, has the criterion: Top importance articles: ``Subject is extremely important, even crucial, to its specific field. Reserved for subjects that have achieved international notability within their field."

Well, if ``the field" is Bayesian statistics, I think it's extremely important, and certainly has achieved international notability. This is probably so even if the field is statistics generally, although in that case perhaps it's not ``extremely important", just important.

In any case, the article badly needs clarification because the definition of conjugate prior is vague (due to vagueness in what's meant by a ``family" of distributions), and the tables, which look to be very useful if accurate, and also look to have received substantial and appreciated attention to improve the accuracy, appear potentially inconsistent with the definition, in that the classes to which they assign prior and posterior distribution are often different, and they don't supply any wider class to which both belong, while earlier in the article, conjugate priors are defined as ones for which the prior and posterior both belong to the same class. Also it seems like this definition must be incomplete because whether the prior and posterior belong to the same class must depend on the likelihood function!

Thanks for any expert attention you can give to this article!

19:45, 21 March 2024 (UTC) MorphismOfDoom (talk) 19:45, 21 March 2024 (UTC)