4

I wrote a speculative paper, outlining an approach to make data analysis more efficient. I am a physicist and the idea is described in words and with a few schemes. I don't have the programming skills to show that the idea works.

In terms of etiquette, would it be OK to publish the paper on arXiv?

  • 9
    A paper on arXiv is a paper. It is linked with your name and reputation, forever. "I don't have the programming skills to show that the idea works." indicates that the idea is not mature enough yet to be a paper. Whether you want to have something like that linked to you, only you can know, but one principle which pops out invariably is: your readers are not your beta testers. – Captain Emacs Apr 23 '17 at 21:54
  • 2
    Have you tried contacting someone who does research in data analysis and programming to see if you can run your ideas by them? They might be able to point out what's wrong (if anything), what about it already exists (could be a lot since you're not an expert), practical problems (there are a lot of obscure arcana to programming due to languages and hardware, such as cases where it's best to start counting at a very large number, instead of at 0 or 1), or suggest a joint work if it seems meritorious but in need of a specialist. – zibadawa timmy Apr 23 '17 at 23:17
  • 1
    I don't have the programming skills to show that the idea works. — Then what makes you think it does? – JeffE Apr 24 '17 at 11:28
  • Does it have any analytical approach? – Mithun Apr 26 '17 at 6:40
  • Nope, it's more like a review/outlook paper, more proposal-like than analytical. – albus_c Apr 27 '17 at 9:21
5

In general when is "I have a general interesting idea for solving problem X but don't have tools Y to determine if it really works" worth publishing? It depends on how important the advance would be, how big a step your paper takes relative to the part remaining, and how easy it would be for the remaining work to be done. I agree with Captain Emacs that you should not be using readers as your beta testers. Let me propose three categories:

A) Not worth publishing: Is (a+b)^2 = a^2+b^2? Simple question, which could be resolved by actually doing the algebra.

B) Probably not worth publishing without extra work/collaboration: Our mathematical model predicts cancer correlates with gene X. Potentially important result, but this is something a bioinformatics person would know or be able to find out easily - so why are you spending so much time writing this up without asking someone to check first?

C) Worth publishing: "The Taniyama-Shimura conjecture, if true, implies Fermat's Last Theorem" (https://en.wikipedia.org/wiki/Ribet%27s_theorem). Linking two unsolved problems, or providing a potentially new attack on an old problem, can be a useful contribution even if it doesn't solve everything.

I think you are probably in case B, because if the issue is just that you don't have many programming skills, you can either learn enough to implement your ideas or find someone to collaborate with. If you can't, this means your question is probably not well-posed enough to interest the broader audience of arXiv. I suspect this is the case, because your question is not too detailed "Data analysis" is a huge field -- you could be trying to speed up an algorithm for a known process (something where you could have a theoretical result -- but it shouldn't be too hard to implement to show you are right), or find a better way of extracting some detail from data (need to test on real-world data to make sure you are not fooling yourself), etc. Your question doesn't sound like you need super-specialized skills, or a giant investment of time to fill in the later details -- so you should try to do this before publishing. Otherwise, you run the risk of spending weeks writing up a draft to wait a month for someone to test your idea to find out that, if you asked the right person, they could have found out in ten minutes that it didn't work.

2

I encountered a similar issue with my dissertation, where I had to develop and prove an algorithm.

My advisor and department chair instructed me to not post the MATLAB demo code I had written and originally included as an appendix, out of fear that (a) it could be stolen (we're on the ProQuest online posting bandwagon for dissertations), or (b) that someone could implement it incorrectly, assume it doesn't work, and tag both me and my new alma mater for low-quality work.

Instead, my graduate committee wanted me to shift focus on providing demonstrative proof that the program both halts (meaning, of course, that it takes a finite number of steps to respond after input) and properly completes its assigned task - in my case, through proof by induction on key loops.

Though I know publishing is different than dissertation writing, I would assume the same standard may be in play here, unless the context of your problem itself takes focus off need for formal proof.

  • out of fear that (a) it could be stolen — This is an argument in favor of posting to arXiv (or better yet, github), since that provides clear evidence of priority. – JeffE Apr 24 '17 at 11:29
  • JeffE, point well taken. – Thomas Rasberry Apr 25 '17 at 0:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.