Re: facebook for stan etc

Hi, I’m only calling it a “facebook” because someone else was calling it that, and this somehow became the name of the project.

Thanks for the detailed questions! See responses below.

  1. I’m not sure what you mean by “link to a discussion offline” so I guess this feature (whatever it is) probably isn’t critical to what I’m interested in.

Could the files live on GitHub and get submitted through
a GitHub pull request, and then we host the discussion of
that file pack on Discourse, where we have our dev discussions
now, and where I’m going to reply.

  1. Logins are fine with me. After all, you have to log in to use Arxiv!

  2. I’m happy with people using their Github logins.

  3. I’m imagining it would have several steps as in my submission:

  • title
  • author 1 name, email, and affiliation
  • option to add more authors
  • acknowledgments (can be blank if there are none, but I’d make it a required field that people would need to choose to leave blank if that’s what they wanted to do)
  • references (can be blank if there are none, but I’d make it a required field that people would need to choose to leave blank if that’s what they wanted to do)
  • abstract
  • Stan file 1 name and description
  • Data file that goes with Stan file 1 (that was not in my sample submission but I was convinced by you and Mike that this would be a good requirement): this could be blank because it’s possible to have a Stan model with no data, but I think a blank data file would be rare in practice.
  • option to add more Stan files
  • option to add one or more scripts in a programming language. For each, user would input the filename (it would be an ascii file) and the language (R, Python, Julia, Stata, Matlab, or other, I guess?)
  • story (can be blank if there are none, but I’d make it a required field that people would need to choose to leave blank if that’s what they wanted to do)
  • challenges (can be blank if there are none, but I’d make it a required field that people would need to choose to leave blank if that’s what they wanted to do)
  • tags (I didn’t have these in my sample submission but I’m thinking they’d be a good idea)

OK, so a big form with validation.

  1. I’m assuming that, yes, users can submit plaintext files in any form but we would not run these files, they’d just be for the user to post. Maybe we should also allow pdf or hrml too, just cos someone might have a clean description that we’d want to upload. I’m guessing that our server would not be in the business of checking or running R/Python/knitr etc code.

Then we need to get every submission to come with
dump file inputs we can run on CmdStan if we want to run
them. Then we need a machine with the cylces to run them.
Or we just save them and don’t run them. We’re not going to
validate they do anything else.

But you think we need all this auxiliary stuff along with
the submission like R programs?

  1. At some point, some sort of markdown could be helpful but right now I’m imagining no, as it seems like one more thing to worry about. I don’t feel strongly about this one.

So do we have a page for each submitted program or each submitted
batch of programs? And it just has the file lists and the text that
came with the form, right?

  1. I’m not sure. Perhaps just a big pile in Github that could be indexed by title, author, submission date, or tags. If we do get a bunch, some organization could be good, but that’s not my first priority.

So just a chrononlogical list, maybe by title and author?

  1. I’m thinking this would be like Arxiv, that users can go back and update their submissions, and then all dated versions are accessible.

OK, so that makes showing everything harder, because we have
to have a way to both store and retrieve and display history.

  • Bob

I would say check that the .stan file parses but don’t check that it compiles or runs.