# Has any methodological research begun with a CrossValidated question?

Are there any publications that were motivated by a CrossValidated post? Was the site acknowledged? Is there a badge for that?

• Not sure about methodological research. Another point is that CV is quite young for any paper to have made it through a publication cycle. Also, for the most part, the questions here are usually not research level, even in applications, as far as I can tell. There are some questions and answers here or there that have likely gotten someone over a hump (or at least out of a trough) on some of their work. Nov 3, 2011 at 23:36
• I think it would be good if someone could write up a paper describing the site, which people who want to acknowledge/thank the site could then reference. Biostar has done something like this. Nov 6, 2011 at 15:59

I thought that the answers to the question "For which distributions are the parameterizations in BUGS and R different?" would be of general interest to the R community so I wrote it up (in collaboration with the answer's author). The article "Translating Probability Density Functions: From R to BUGS and Back Again." is methodological, though the research part was limited to compiling references.

The acknowledgements read "This collaboration began on the Cross Validated statistical forum (http.//stats.stackexchange.com/q/5543/1381)."

Yes.

Our paper summarizing this discussion: Is ridge regression useless in high dimensions ($n \ll p$)? How can OLS fail to overfit? has just been published by The Journal of Machine Learning Research:

The Optimal Ridge Penalty for Real-world High-dimensional Data Can Be Zero or Negative due to the Implicit Ridge Regularization
Dmitry Kobak, Jonathan Lomond, Benoit Sanchez; 21(169):1−16, 2020.
https://jmlr.org/papers/v21/19-844.html

It was a preprint since mid-2018: https://arxiv.org/abs/1805.10939 but took some time to publish formally (and got substantially extended over the course of revisions). JMLR is a very respectable journal, so I am happy :-)

We of course acknowledge CrossValidated:

@amoeba, @JonnyLomond, @BenoitSanchez

• Very respectable indeed! They get the first opportunity to reject all my papers. Aug 26, 2021 at 17:03
• I forgot to say, congratulations! Aug 26, 2021 at 17:17

Hopefully there will be at some point. I am performing an empirical study of Nelder-Mead simplex, gradient descent and grid-search methods for model selection for kernel machines, which was inspired by a question on Cross Validated (which appropriately is also the model selection criterion ;o). No paper yet though.

• The paper is currently in review. Jul 18, 2013 at 12:04
• Has this paper been published by now? Oct 22, 2015 at 15:47
• No, it was rejected (despite statistically significant result ;o), am still working on the second draft, hopefully it will be a better paper for the revisions. Oct 23, 2015 at 8:28
• Did you ever get this published?
– Dave
Jan 20 at 22:40
• No, unfortunately I have had some issues in real life that have limited my research activity (and a huge teaching load!) It is quite difficult to get rigorous studies of things that make minor, but worthwhile improvements to best practice. Especially frustrating as what I was suggesting is less computationally expensive! Jan 21 at 9:28

Almost two years after having asked this question, I actually wrote a paper inspired by CV (still unpublished):

http://arxiv.org/abs/1304.4920

I used the question's url in the body of the text.

• John could you correct the name of the site in the title and body of your post here (you're missing a "d") and also in your paper (replace the terminal "s" with a "d")? Jul 19, 2013 at 2:49
• Thank you. I edited this post. Could you be more specific about the changes needed in the paper? Jul 20, 2013 at 12:37
• In the abstract, fix the name of the site - i.e. the same word as you corrected here. That is, taking the abstract from here: "in the statistics Questions & Answers web site Cross Validates", replace the terminal s with a d. You could just search for 'cross'. Jul 21, 2013 at 0:01
• This issue (that the summary at the link refers to our site as "Cross Validates") remains. Sep 20, 2020 at 6:57

A badge for it would be tricky, as it's not an automatic thing that can be scraped from the site itself.

I don't know of any methods research spawned by CV, but I haven't been here all that long, so someone may come and chime in. I have however used it to sanity check an idea or two that I've had. Though the one time CV did out-and-out answer the question, I had a better source provided.

I'm currently looking into the class imbalance problem, which is a very common methodological issue on this site, and have been working on a tutorial paper. The lack of an answer to my question "How do you know that your classifier is suffering from class imbalance?" suggests there are some issues with current practice that need to be addressed. Cross Validatated seems a good place for finding out what people actually do (or don't do)