StackExchange in general and this Signal Processing group in particular have been very useful in helping us use fft for a research project. However, we're trying to publish this stuff in a scientific journal, and I'm expecting some concerns from reviewers about citing a source like this. Any suggestions? EndNote has a reference type "Web Page", so we'll use that format, but any suggestions on how to defend them as comparable to more traditionally peer-reviewed sources?
SE.DSP is peer reviewed to some extend. Not in the traditional "journal" sense, meanwhile predatory publishing or randomized peer-review, to name a few, have changed the deal. A post in SE.DSP can sometimes be much more peer-reviewed than a paper in some conferences/journals.
However, SE.DSP as a site has no DOI. It can vanish, questions and answers can be edited, deleted. There is no means, as far as I know, to freeze a citable version, as you could do for other web sources like wikipedia. Suppose that you cite a page for a "true result", to sustain a part of you paper. If it changes, turns wrong, or disappears, your results may remain correct, but without a source. In that sense, SE is not auditable at the time of writing. As said in Citing References in Scientific Research Papers:
It is important to properly and appropriately cite references in scientific research papers in order to acknowledge your sources and give credit where credit is due.
And about the sources:
Sources that need to be acknowledged are not limited to books and journal articles, but include internet sites, computer software, written and e-mail correspondence, even verbal conversations with other people (in person or by telephone). All different kinds of sources must be acknowledged.
So I doubt a Q&A link can be used directly, as of today, for a "scientific publication". But... a publication is about making something public. It does not anymore restrict the sole paper pdf. So I propose these options:
- if you want to pay credit to SE.DSP, vote answers up, share the OP, put a link to the related answer on a blog, webpage, etc. together with a link to your paper; you can even refer to the main SE.DSP site as a source of inspiration;
- if you want to pay credit to someone, ask him for the reference, thank him in the paper, add a "personal communication" to the reference list. But you likely are to rephrase the mentioned result in your paper, so that it can be checked. You can even collaborate with someone;
- if you want to pay back: add your answer, contribute the community wiki. Perhaps one day our community wiki will be truly citable;
- [EDIT] As proposed by Marcus Müller (credits well deserved), one could resort to forms of digital archive. This link to Citing from a Digital Archive like the Internet Archive: A Cheat Sheet proposes techniques to do that. There was also a quora question on How should I cite an archived version of a web page in APA style?
Ok, though I really think Peter, rbj, Matt, Laurent, Rick et al looking over my answers here is a much better review process than, let's say, a lot of IEEE conference proceedings use, it's really much less work to post an answer here.
Also, I do know I make a lot of mistakes; if I was writing a paper, there wouldn't be a single non-basic "from the top of my head" formula in use. The content/effort tradeoff is just a lot more in favor of content on here.
Really don't cite a StackExchange answer, if at all substitutable with proper publications¹. Most of the things said here are really meant to incite looking things up yourself. For example, there's a high probability that if I think the original poster of a question is a beginner, that I'll give them a rough idea of how things work, and put single google-able terms in italic to inspire research.
If I think there's something that deserves a quotation (because, for example, I know of the common source that everyone refers to, example Schmidl&Cox, example Bellanger, example harris, example Tukey), I often just actually cite – something that is totally optional here, but would be a severe mistake if omitted in scientific literature.
If I wrote something like
the FFT is just a specific implementation of the DFT, which you can understand as a time/frequency domain transform, but also just as a square, invertible matrix operation that maps $\mathbb C^N \mapsto \mathbb C^N$.
in a tone of "I think I can just state that without much explanation", you can be relatively sure that I'd expect to find this in a common DSP textbook. Just ask the author of an answer if you can't find a source in a few popular ones at your local library; she/he might have a favourite book to point to.
Also: We're all lazy. I certainly don't have all formulas I need in my head – but I know which Wikipedia articles to read to either find the formula (be very careful with wikipedia itself, though, if you've never seen the formula itself before. There's a lot of plain wrong math on there.) or at least proper sources (wikipedia is usually very stringent about citations).
If I add a derivation that is really detailed (like: Ten lines that actually show how to transform a not-that-complicated equation), this might indicate that this is really basic stuff, and you should definitely pick any textbook and start reading it from the start until you hit the topic of interest.
Researching "proper" sources is part of the scientific method. It's often pretty boring, but it reduces the risk of accidentally citing something that has a small arithmetic error, or a misdirected train of thought.
Many of us know that problem and will happily share their sources.
If you really can't find a source for something, and the poster of an answer doesn't react, don't feel bad about asking a question among the lines of
Where is foobaring described? I saw it in answer link, but I haven't been able to find a proper source. I've looked into Prosklar's Digital Fooing and Kammharris' Foundations of Signal Baring, but all they introduce are the basics. Google Scholar turned up Shaquist's paper On the limits of foobaring on noise-limited schmerxels, but that just uses the concept as given, and doesn't cite any source that I'd be able to read.
and tag it with reference-request.
¹ This actually depends a bit on the author of the answer. If you happen to know a specific answerer is highly reputable on his field, well, that might or might not be sufficient for whoever reads your paper. I certainly wouldn't cite Müller, M.: Proceedings on the DSP StackExchange without a lot ado, whereas Lyons, R.: On the very nature of DSP, Signal Processing StackExchange might have a different ring to it in the ears of the reader.
However, I'd say that might be best-case applicable for Proceedings in general (there are certain things that you can even cite as "private conversation with XYZ", but that depends on both things and XYZ); Journals (if they are worth anything) will have a review process that will expect you to find easier-to-convince-people-with sources.
Stackoverflow answers can be edited and even deleted at almost any time.
So to cite a stackexchange answer, you probably need to find a trusted reliable archival site that can (and has the rights to) take a frozen snapshot in time, and cite that snapshot as well as the originating site.