Anonymous R packages on CRAN or Github to accompany blinded peer review of submitted manuscripts [closed]

Question

Is there a way to share R-package anonymously that will work on Unix, Mac, and/or Windows (preferably all 3 and with the ease of having it on CRAN)?

Having an R package on CRAN so that analyses can be reproduced and methodology can be demonstrated and shared can be a big boost to the review of a manuscript submitted to a scientific/statistical journal (in my opinion and experience).

If that journal requires blinded reviews, how can I share the R package in a way that keeps the review blinded (traditionally, the DESCRIPTION file lists my name and email address, which would unblind the review)?

I have thought of the following options, all of which have drawbacks:

1. Go through whole CRAN submission process with a pseudonym (fake name and throwaway email account) without using github (my github username is my last name). After the review is unblinded / paper is accepted change the throwaway information to the correct information. I'm uncertain of the etiquette of this or how CRAN's policy would deem this practice.
2. Zip up R package with no involvement of CRAN or Github and trust the reviewer is interested and capable enough to install it from source on unix. There's a big difference between this and being able to type install.packages() and library() on the system the reviewer is familiar with, and manually creating and including zips for all platforms is tedious.
3. Don't make a package, just send code snippets and data and state in the manuscript an R package is forthcoming (which is a weaker statement than, "here's the R package that is already on CRAN"; another drawback is that listed in item 2).

I've mentioned CRAN and Github because I'm most familiar with these repos. I'm open to other solutions.

Answer 1

There’s no need at all to have the package on CRAN,¹ and there’s no way to submit packages anonymously to CRAN. Such a submission would be a big problem for CRAN in terms of maintainability. CRAN is simply not the correct platform for this.

Github has similar issues but in principle you could just create a separate Github account without providing identifying information.

However, this just sidesteps a bigger issue: How non-identifiable is your code really? More generally, the whole idea of double-blinded peer review is dogged by issues of identifiability of the research. I don’t think there’s a good solution (especially involving code review, but even in general) where the research is submitted anonymously. As such, I don’t think it’s worthwhile spending energy trying to make code submissions anonymous, to the detriment of software (maintenance) quality.

In cases where double-blind anonymous peer review is desired, the currently best option is to submit the code to a service that allows anonymous archival, such as Figshare, or submit an archive as supplementary material to the journal. It should not be a stretch to expect the reviewer to perform a simple

install.packages(path_to_file, repos = NULL, type="source")

… otherwise they may not be qualified to review the code anyway.

---

¹ In fact this isn’t even desirable (on the contrary, I find the cluttering of CRAN quite counter-productive; though “CRAN” has “comprehensive” in its name, ideally all its contents should be in the form of properly usable packages; in other words: quality, not quantity).