Wikipedia:Articles for deletion/Probit model for panel data with heterogeneity and endogenous explanatory variables (2nd nomination)

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. Lourdes 00:15, 27 October 2019 (UTC)[reply]

Probit model for panel data with heterogeneity and endogenous explanatory variables[edit]

[Hide this box] New to Articles for deletion (AfD)? Read these primers!

(Find sources: Google (books · news · scholar · free images · WP refs) · FENS · JSTOR · TWL)

I don't know how this one survived the first AfD nomination. It is hopelessly over-technical; something that (if at all) should be mentioned in a section in Probit model. Right now this is the econometrics equivalent of having an article iPhone 11 in yellow and with 256 GB storage on top of iPhone 11. bender235 (talk) 02:31, 17 October 2019 (UTC)[reply]

Delete per nom, clearly way too technical. Ten Pound Hammer • ^{(What did I screw up now?)} 03:01, 17 October 2019 (UTC)[reply]

Keep. I concur with User:Mark viking's "Keep" !vote in the first AFD. --Doncram (talk) 03:07, 17 October 2019 (UTC)[reply]

Mark viking's comments seem beside the point in my opinion: nobody doubts that this particular model "is real and seems adequately referenced." It sure is, but it's an overly special corner case that does not merit its own article. Next thing we have an article on Tobit model in an unbalanced panel with interaction terms and so on. --bender235 (talk) 04:01, 17 October 2019 (UTC)[reply]

Comment - Neither "Too technical" nor "Too detailed" is a WP:DELREASON. What actual policy reason is being invoked here? Is it WP:NOTGUIDE? FOARP (talk) 14:38, 17 October 2019 (UTC)[reply]

I would say it's WP:INDISCRIMINATE. Basically the same reason why we don't have the article iPhone 11 in yellow and with 256 GB storage. --bender235 (talk) 15:24, 17 October 2019 (UTC)[reply]

I get what you're saying here, but I think WP:REDUNDANTFORK might be a better fit - we don't have an article on iPhone 11 in yellow and with 256 GB storage because its simply a redundant fork of the iphone article. Alternatively you could say it's a WP:POVFORK because it's giving WP:UNDUE emphasis to relatively minor aspects of the iphone. FOARP (talk) 11:24, 18 October 2019 (UTC)[reply]

Note: This discussion has been included in the list of Mathematics-related deletion discussions. XOR'easter (talk) 20:38, 17 October 2019 (UTC)[reply]

Delete. Written so increadibly badly that the article is incomprehensible even to the experts, so much so that the page is a borderline WP:A1 case. I am a professional mathematician myself and I have no idea what the page is talking about. There are no definitions of any kind, the topic of the article is not coherently described, the various quantities/notations/variables used are not defined or explained, and there are no clearly idetifiable and properly formulated mathematical statements. Basically just a bunch of incoherent semi-mathematical verbiage. In mathematics we refer to this kind of text as "not even wrong". Nsk92 (talk) 21:43, 17 October 2019 (UTC)[reply]

Well, I am an econometrician and can attest that the article is correct, although poorly written. However, even the most beautiful rewrite does not change the fact of my opening statement, that this particular model is too much of a special case to merit its own article. --bender235 (talk) 23:51, 17 October 2019 (UTC)[reply]

The article is no more "correct" than the sentence "If

x_{ij}+U_{k}^{i}=42

then

\int _{\pi }^{c_{ji}}V_{mn}\otimes e^{\alpha }dx=0

." — Preceding unsigned comment added by Nsk92 (talk • contribs)

I happen to have a very clear understanding of the basic probit regression model, and I am sure I could explain it (teach it) a lot better than is done in the current probit model article. About heterogeneity, and/or about endogenous explanatory variables, I could not do that. I am more in the category of wishing that this article were far better/clearer than it is, so that I could understand it properly. This article should be written at level comprehensible to someone who has taken coursework about equivalent to that required for a masters degree in statistics, I think. Perhaps this article should be split into different topics, or revamped considerably to cover multiple refinements of the basic probit model including heterogeneity of errors, endogeneity, more variations. I cannot say that editor Bender235 is wrong, in their estimation that the variation(s) addressed in this article are relatively obscure. But I think the solution would be to revise and to expand this, and re-title it as appropriate. Probably there should be one article (this article should be revised to become) on variations upon the basic probit regression model. Sticking to binary responses, setting aside extension of the basic model to multiple levels of responses. --Doncram (talk) 15:29, 18 October 2019 (UTC)[reply]

Question Would anyone with experience in stats or econometrics be willing to put in the time to clean this up so it might be merged, e.g. to probit model? There's also Binary response model with continuous endogenous explanatory variables, which suffers the same difficulties as this page: too obscurely written even for people who know math, and devoted to a case of narrow scope. XOR'easter (talk) 04:26, 18 October 2019 (UTC)[reply]

That sounds very reasonable to ask if someone with more experience than I have, whether they could fix up this article better. As i explain in comments above and below, I don't think this should ever be merged into a lower-level article on the basic probit regression model. --Doncram (talk) 15:29, 18 October 2019 (UTC)[reply]

Delete - I think this article fails a number of points on WP:NOT. It is a text-book style article and as such fails Wikipedia is not a text book. It also appears to be a how-to style article and thus fails WP:NOTMANUAL. It is not written in accessible language and thus fails the "no academic language" requirement. Moreover, whilst I see a number of supporting references in the article, all I can see are the titles, so I do not know if they directly support the notability of this topic, so I have concerns about the extent to which WP:GNG is actually met - whilst I am assuming the good faith of the editors who added it to the article it would be good to have this confirmed by someone with access to these journals. Also, this is a WP:REDUNDANTFORK or even WP:POVFORK (since it is avoiding WP:UNDUE) of Probit model. FOARP (talk) 07:09, 18 October 2019 (UTC)[reply]

FOARP has valid complaints about the article, which would properly be addressed by editing the article, not deleting it, IMO. But it is not redundant or POV in any way, relative to the basic probit model. --Doncram (talk) 15:49, 18 October 2019 (UTC)[reply]

Question - In the previous discussion, another user doncram indicated that it would be inappropriate to merge this article with probit model. If that's true, then can this doncram (or anyone else with expertise) provide a clarification for why the content could not be included as a subsection of that article? The article itself is not especially long so it's not a space issue , but if there's a technical reason why presenting this information as a separate subsection under that main heading is inappropriate then it may be best to not merge it and to leave it as a standalone article as now with an explanation for why it's kept separate documented in the talk page. 208.185.237.210 (talk) 13:23, 18 October 2019 (UTC)[reply]

Sure, that's easy. This article cannot be merged to the probit model article because this topic is too complicated. The "probit model" article actually should be simpler than it is. For example, it should have graphics explaining the model much more clearly (I wish i could easily provide those... I could hand-sketch what is needed, but I currently can't generate good 3-D graphics). It probably should have one or maybe two well-chosen example applications. Mainly, it should be written to be accessible to persons having a basic understanding of statistics, such as from just one statistics course in college or just one "Data and statistics" type course in business school, say. It should clearly lay out the idea of probit regression following from a few "simple" assumptions. The ideas of heterogeneity and endogeneity are far, far, far too much for the first, introductory article about probit regression.

If this article were merged with the probit model article, then the only sensible options, for editors of that article, would be to split the merged material right back out again, or entirely delete it all. --Doncram (talk) 15:11, 18 October 2019 (UTC)[reply]

Thank you for the clarification. With that additional context, it does seem logical to keep this article as it is and, probably on the talk page, discuss trimming it down so that it is at the appropriate level of language. It might also be useful to have an external link pointing to this article in probit model, similar to how we have related articles point to each other. 208.185.237.210 (talk) 15:57, 18 October 2019 (UTC)[reply]

You're welcome. :) Yes, a "See also" type link at the bottom of that page is appropriate. (It wouldn't be labelled an "External link", because those go to locations outside of Wikipedia.)

Further, in the basic probit model article there is a section "Performance under misspecification" which should be removed from there, and put into a more advanced-level article. I suppose that heterogeneity and endogeneity are special cases of misspecification of the basic probit model. I think this article should be modified to cover a number of variations, each really being ways to address various misspecifications of the basic probit model. User:bender235 somewhere above suggests, perhaps facetiously, something about covering "unbalanced panel with interaction terms" stuff; maybe that is one more type of thing to be covered as a variation? Also the normality of distribution of errors is a strong assumption which should be relaxed; the advanced article should at least suggest how assuming heavier-tailed distributions is probably almost always better (because in any real application there are no doubt unmeasured, omitted explanatory variables, which in effect cause greater variation in outcome than would be expected in the simple model) and can actually be easily addressed... logit regression is one alternative different only in this way. Assuming Cauchy/other stable distribution function are options, maybe adding one or two more parameters to be estimated. Offhand, I am not sure how to rank the importance of the various variations. --Doncram (talk)

Just for the record, because it seems this was misunderstood: I suggested "something about covering "unbalanced panel with interaction terms" stuff" as a joke. I do not want this kind of an article. Neither do I want iPhone 11 in yellow and with 256 GB storage. --bender235 (talk) 18:20, 18 October 2019 (UTC)[reply]

Also, by the way, the basic probit model article should not have to cover Berkson's whatever and Gibbs sampling, which I think might both be considered computational/solution aspects. Computation/solution method stuff should probably be covered in a different advanced-level article. About computation, the basic article should include just the assertion that the maximum likelihood surface is convex and has just one maximum to be found (i.e., under the assumptions of the probit model, for any one data set there is just one "best-fitting" set of coefficient estimates, and these can be found by a simple search process ...there are no local maxima to worry about), and brief suggestion of one simple approach to finding way to the maximum should be given there. --Doncram (talk) 16:51, 18 October 2019 (UTC)[reply]

- Delete Way way too niche. All the creations ought be nuked per User_talk:Carolineneil#Single_purpose_account_for_experiment_on_Wikipedia:_should_be_banned and the subsequent block. ∯WBG^converse 16:52, 18 October 2019 (UTC)[reply]
Strong keep - It's clear that some of the contributors to this discussion lack subject matter expertise on this topic, but to clarify this topic cannot be comfortably folded into probit model, though it is notable enough to support a standalone article. 107.77.203.224 (talk) 17:54, 18 October 2019 (UTC)[reply]

I do have expertise in this area, but as FOARP correctly pointed out, Wikipedia is not a textbook. We discuss topics like the probit model for a general audience, and leave specializations like Probit model for panel data with heterogeneity and endogenous explanatory variables for textbooks. --bender235 (talk) 18:22, 18 October 2019 (UTC)[reply]

Delete per bender235.4meter4 (talk) 16:23, 24 October 2019 (UTC)[reply]
Delete, as non-encyclopedic. If the base article needs to be easy to understand, so does this. - ChrisWar666 (talk) 19:24, 26 October 2019 (UTC)[reply]

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.