Template talk:Statistics

From Wikipedia, the free encyclopedia

Contents 1 Statisticians 2 More topics for navbox 3 Link to project 4 Split? 4.1 Econometrics 4.2 Experimental design 5 Sample size estimation 6 Collapsing

[edit] Statisticians

Should the template include notable statisticians? I think it may be too much, but what does everybody else think? Lilac Soul ^{(talk • contribs • count)} 08:23, 28 August 2007 (UTC)

If we do, we should probably only include the top five or six or so. --Arcadian 11:47, 28 August 2007 (UTC)

It doesn't seem relevant to me. The purpose of the template is to make it easier to navigate to related topics, and I think it's unlikely anyone will want to navigate from t-test to R. A. Fisher. --Zvika 11:37, 9 October 2007 (UTC)

It sounds like a good idea to me, but perhaps it could be it's own navbox?

[edit] More topics for navbox

Could topics like Bimodal distribution and Binomial distribution be included? They seem to be already included in the ProbDistribution navbox, but that navbox is very large. Rhetth (talk) 22:41, 4 August 2008 (UTC)

[edit] Link to project

Apparently, links to Wikipedia:WikiProject_Statistics in this template violates WP:SELF - WikiProjects should not be included in articles or on their templates. I have removed.—G716 <^T·_C> 01:13, 22 September 2008 (UTC)

[edit] Split?

This template has become quite large, and as WP:NAV says, "a large template has limited navigation value". I suggest splitting it into several smaller templates on each sub-topic. (We could draft them as sub-pages of this talk page to start with, e.g. Template talk:Statistics/Experimental design.) We could also consider a "Major topics in statistics" navbox to give an overview, in the same style as e.g. Template:Number theory-footer, Template:Applied-footer etc. Thoughts / volunteers? (I'm not taking this on on my own..) Qwfp (talk) 10:22, 1 October 2008 (UTC)

"Quite large" is, of course, relative. This template is bigger than a breadbox ({{Mathematics-footer}}), but not as big as a Mack truck ({{HIVpharm}}, {{Medicine}}, {{Influenza}}, {{Compression Methods}}). Personally, I think that it's ok at the moment, but I would support removing the "History" section (that I added by don't own) and the "Publications" section to keep the template from becoming list cruft. I'm not sure that some of the possible sub topics are big enough to rate a separate template. If this idea moves forward, I'll be willing to help out. Regards—G716 <^T·_C> 01:04, 2 October 2008 (UTC)

We could make it collapsible: {{Herbs & spices}}—G716 <^T·_C> 01:12, 2 October 2008 (UTC)

or collapsible, with autoshow groups: {{ProbDistributions}}—G716 <^T·_C> 05:57, 3 October 2008 (UTC)

We could use {{Least squares and regression analysis}} as one of the sub-topic templates.—G716 <^T·_C> 05:49, 3 October 2008 (UTC)

I have removed publications and history sections to reduce size of template. If anyone objects, please revert. —G716 <^T·_C> 12:20, 16 April 2009 (UTC)

[edit] Econometrics

Perhaps have a separate template for the subset of statistics that pertains to econometrics? cancan101 (talk) 02:19, 12 March 2009 (UTC)

I don't think there is a subset of statistics that pertains only to econometrics. But that doesn't preclude having an econometrics Nav template as well: I don't know of there already is one. Melcombe (talk) 09:21, 12 March 2009 (UTC)

[edit] Experimental design

Following the previous discussion, I augmented the template on design of experiments, resulting in the operational template Template:Experimental design. The template is designed for (somewhat) more knowledgable users, than is the basic statistics template. For example, Student's t is listed next to Hotelling's T, to provide context for the latter.

v • d • e Design of experiments Scientific Method Scientific experiment · Statistical design · Control · Internal & external validity · Experimental unit · Blinding - Optimal design: Bayesian Random assignment · Randomization · Restricted randomization · Replication versus subsampling · Sample size  · Power · Type I & II error Treatment & Blocking Treatment · Effect size · Contrast · Interaction · Confounding · Orthogonality · Blocking · Covariate · Nuisance variable Models & Inference Linear regression  · Linear model · Best linear unbiased estimation (BLUE)  · Linear least squares  · Maximum likelihood · Bayesian Random effect · Mixed model · Best linear unbiased prediction (BLUP) · Restricted maximum likelihood (REML) · Hierarchical model: Bayesian Analysis of variance (Anova) · F-distribution & Test · Chi-square distribution  · Cochran's theorem · Manova (multivariate)  · Ancova (covariance) Test · Most powerful · Paired · Compare means · Multiple comparison · Omnibus Nonparametric tests: Permutation · Exact · Wilcoxon signed-rank · Mann–Whitney U · Kruskal-Wallis · Friedman Designs: Completely Randomized Factorial · Fractional factorial · Plackett-Burman  · Taguchi Response-surface design · Polynomial & rational modeling · Box-Behnken · Central composite Block  · Latin square  · Graeco-Latin square · Hyper-Graeco-Latin square · Latin hypercube Repeated measures experiment · Crossover experiment · Missing & Censored data Randomized controlled trial · Sequential analysis · Sequential probability ratio test Glossary · Category · Statistics portal · Statistical outline · Statistical topics

A few Bayesian entries occur at the end of the lines (which provides some vertical organization, also)
The horizontal line between inference and designs is awkward, but should aid the reader (imho). (Maybe somebody knows a better solution?)
Feedback now is especially welcome.Kiefer.Wolfowitz (talk) 11:49, 23 June 2009 (UTC)

[edit] Sample size estimation

Any thoughts on the relevance of the first three items under this heading: Null hypothesis • Alternative hypothesis • Type I and Type II errors • Statistical power • Effect size. I would suggest dropping them entirely as they aren't needed elsewhere in a brief list of topics. Melcombe (talk) 15:55, 9 December 2008 (UTC)

Type I & II error needs to stay, because reducing the probability of a Type II error is the usual concern for minimizing the number of replications (for ethical & economical reasons). Also, the Type I error is usually fixed; the phrasing is "Type I & II error", so keeping Type I error requires only 4 characters, "I & ". Removing the hypotheses is okay. (Duplicates comments at the ExpDes Template, addiing the character count ) Kiefer.Wolfowitz (talk) 15:45, 25 June 2009 (UTC)

On this template's talk page, I provided an example of a substantially shortened experimental design template. I commented out the topics already suggested for deletion. I commented out the RSM geometries (cube, simplex, sphere), the algebra of balanced designs (association schemes, Bose Mesner algebra, spherical designs), and the specialty entries on sequential designs (armed bandits, gittins index). I put the observational studies topics on their own line (below repeated measurements, because of similarities and the weakness of the Repeated measurements article, imho).

[edit] Collapsing

The option {{Statistics|state=collapsed}} appears not to work .... can it be fixed? Melcombe (talk) 10:21, 2 December 2009 (UTC)