Wikipedia:Articles for deletion/Distance-weighted estimator


 * The following discussion is an archived debate of the proposed deletion of the article below. Please do not modify it. Subsequent comments should be made on the appropriate discussion page (such as the article's talk page or in a deletion review).  No further edits should be made to this page.

The result was   delete. It seems that the concept is not yet sufficiently widespread to have received enough coverage for Wikipedia - when such coverage arises, the article can be recreated. Yunshui 雲 &zwj; 水  11:52, 27 November 2012 (UTC)

Distance-weighted estimator

 * – ( View AfD View log  Stats )

Original research, self-advertising

Russian section of Wikipedia is currently discussing this subject. It is in fact described nowhere except for the paper written by scientists who suggested this statistical measure, and we have some ground for suspicions that this article had been created by Yury Dodonov himself. I find that this article represents original research and implies self-advertisement, so I suggest that it should be deleted. --Andiorahn (talk) 13:35, 12 November 2012 (UTC) — Andiorahn (talk&#32;• contribs) has made few or no other edits outside this topic.


 * Comment Since the article is based on a published paper, I don't think the article itself contains original research. On the other hand, I am a new editor and may be missing something. The article has no other independent references and may fail notability guidelines because of it. It seems a well written article. I defer to my Russian colleagues and others for recommendation of keep or delete.Mark viking (talk) 22:44, 12 November 2012 (UTC)
 * Note: This debate has been included in the list of Science-related deletion discussions. • Gene93k (talk) 00:37, 13 November 2012 (UTC)


 * Delete. There is no evidence that the topic has received the widespread attention needed for notability. Xxanthippe (talk) 01:27, 13 November 2012 (UTC).
 * what is the problem? really good measure keep Ging72 (talk) 02:11, 13 November 2012 (UTC) — Ging72 (talk&#32;• contribs) has made few or no other edits outside this topic.
 * "Good" is not enough. Extensive sources are needed. Xxanthippe (talk) 02:52, 13 November 2012 (UTC).
 * um… may be. but formal viewpoint isn't always good viewpoint. in my opinion if article will be deleted it will be a formal mistake Ging72 (talk) 05:16, 13 November 2012 (UTC)
 * Never mind. When the hordes of references that the topic no doubt deserves eventuate the article can be recreated. Xxanthippe (talk) 05:21, 13 November 2012 (UTC).


 * This is not original research in the sense intended in Wikipedia's policy; i.e. it does not embody findings published for the first time in a Wikipedia article. I'm currently leaning slightly towards keep. Michael Hardy (talk) 03:59, 13 November 2012 (UTC)
 * Delete! This is coi and undue. There is an extensive literature on this topic, which is ignored: Books on Voronoi tessellations, potential theory, multivariate analysis. David Eppstein has written on related topics. (Alternatively, a stub could be retained.) Kiefer  .Wolfowitz  19:50, 13 November 2012 (UTC)
 * Delete This is an ad hoc formula used in a single paper. We need more than that for an article. If the formula is used in several other papers then the article can be re-created. We don't usually allow academics to publicize their work in this way. I think Wolfowitz means there is extensive literature on measures of central tendency, not on this specific formula or topic. Dingo1729 (talk) 00:54, 14 November 2012 (UTC)
 * Delete. Isn't this just the same thing as the (much more well-established) geometric median? The idea of finding the weighted mean of a set of points, weighted by their inverse distance from the result, seems like exactly how Weiszfeld's algorithm for the geometric median works. So this seems to be a reinvention of the wheel under a new and non-notable name. —David Eppstein (talk) 06:59, 16 November 2012 (UTC) Incorrect — see below discussion. —David Eppstein (talk) 17:24, 16 November 2012 (UTC)
 * No, this is not the same. Not going into details - the geometric median coincides with the median in the one-dimensional case, and the distance-weighted mean (for unidimensional data) is obviously not the same as median. YuryD (talk) 09:25, 16 November 2012 (UTC)
 * You say it's obvious, but it's not obvious to me. In what respect is this different from Weiszfeldt's algorithm? —David Eppstein (talk) 15:45, 16 November 2012 (UTC)
 * Briefly, Weiszfeldt's algorithm minimizes the distance between a point-estimate of location and the actual data points (doing this iteratively), thus producing a median (let's consider only the one-dimensional data). Distance-weighted mean (DWM) has nothing to do with the distance between the obtained estimate of location and the actual data points, it deals only with the distances between each given data point and the other data points to produce weighting coefficients. Am I right that you expect me to show that the algorithm for DWM does not produce a median (which is in fact computed by Weiszfeldt's algorithm), or in other words that the sum of distances between the DWM and sample data points is not necessarily minimal? (for a general case, because the two estimates coincide for a symmetric distribution like many other measures of location) YuryD (talk) 17:11, 16 November 2012 (UTC)
 * Ok, you're right, I misread. Weiszfeld is the weighted mean, inversely weighted by the distance to the center; this one is a weighted mean, inversely weighted by the average distance to the other points. I stand by my delete, but for different reasons now: I can't find any reliable secondary sources that discuss this and put it into context with other central tendancies, so it appears to fail WP:GNG. —David Eppstein (talk) 17:24, 16 November 2012 (UTC)
 * Thanks a lot! Yes, this measure is a new one – we first presented it in 2011, and this was not because of a theoretical interest - we primarily intended to use it for our research purposes (analysis of response times in cognitive tasks). Since that, we published two papers in Intelligence with analyses that were based on this measure (and related measures such as distance-weighted standard deviation and distance-weighted standard score), but it doesn't look like this would qualify for reliable secondary sources (these were practical applications, not even comparative simulation studies). When I created a wiki page on this measure a year ago I simply was not aware of this requirement. That's fine with me if you delete it because of the absence of citation – we can only wait and see if it is extensively used or not. I really did not even intend to show up in this discussion, I only didn't want to end up with the assertion that this was a reinvention of the wheel. Thanks again, Yury S. Dodonov, PhD YuryD (talk) 18:27, 16 November 2012 (UTC)


 * Relisted to generate a more thorough discussion so a clearer consensus may be reached.


 * Please add new comments below this notice. Thanks,  MBisanz  talk 00:04, 20 November 2012 (UTC)


 * Delete puff piece. --Sue Rangell ✍ ✉ 21:30, 20 November 2012 (UTC)


 * Delete After looking over the discussion and searching some more for independent references and finding none, I am changing my neutral stance. I suggest deletion on the basis on lack of independent reliable sources. Except for lack of sources, the article and paper seem like perfectly fine research. The idea seems to bear some relation to Inverse distance weighting interpolation methods. The paper is recent, however, and it is probably too soon for the distance-weighted estimator method to diffuse into reviews, books and news sources. When such independent sources become available, recreation of this article would be reasonable. Mark viking (talk) 22:09, 20 November 2012 (UTC)
 * The above discussion is preserved as an archive of the debate. Please do not modify it. Subsequent comments should be made on the appropriate discussion page (such as the article's talk page or in a deletion review). No further edits should be made to this page.