Social Collaboration Scratchpad

Social Collaboration: Scratchpad

Simple "scratchpad" for organizing topics.

We will be doing two things at once: brainstorming each of the topics, and fleshing out further important topics for a future conference.

• Intro: survey/preview topics for which we will have roundtable discussions. Two main questions here:

  1. How do we incentivize people broadly to use modular code, participate, and give back?

     • Both professional level and graduate students
     • platforms for collaboration
     • Getting citations and peer review for code
     • Collaborative events

  2. How to identify next round of modules/topics needed?

     • "Building the plane"
     • drive everything with examples

• TOPIC: What is list of models we'd like to see, now or in next 9-36 months, produced with a toolkit? [MC: Chris and Nathan]

  • everyone likes wishlists.
  • poll research organizations: "what are current toolkit implementation needs? What would be useful to have?"
  • poll researchers: models/ideas which would be good to see in same framework? (eg. bequests)
  • Garner a list beforehand of models for which we'd like examples
  • Our contribution: two models in progress now; ideas for models which may be useful at CFPB

• TOPIC: Platforms for collaboration? Issues with Uptake? The CFPB experience. [Bill Shelton]

  • What tools to foster collaboration here? What infrastracture?
  • Bill talks about general topics and experiences at CFPB
  • Open-source issues with uptake and collaboration
  • various collaboration efforts -- econforge, quant-econ, replication-wiki. What interactions might look like.
  • [MC: Bill kicks this off]

• TOPIC: How to incentivize professional participation? Options include: [Chris and Nathan]

  • Getting a strong early showing of good models (eg at "public launch" -- quant-econ, stackexchange as good example)

    • A theme will be, "examples drive development" -- replication, new research (incl. dissertations, current research, policy models)

  • A lesson from StackExchange: get people to make specific, time-limited contributions.

    • What might these look like for professional economists? Egs:
      • mentoring commitment, as below
      • posting code: advisers commit to make it a condition of dissertation acceptance that students post code in particular form.
      • commit to posting HW code archives
      • Nominate selves as "gurus" for topics (eg interpolation)
      • Q: what are others?

  • Graduate students are the great "natural resource" here - How to attract/encourage their participation?

    • Advisors
    • Student awards (SCE/CEF)-style awards for particularly useful/desired models
    • Mentoring of model areas by experienced researchers
    • Partner with replication wiki to encourage replication with toolkit

  • Peer review and citations of code -- next topic. "If you cite it they will come."

• TOPIC: Citations and peer review of code [MC: Pablo]

  • example of JSS: open journal with citations for code via "vignettes" posted to CRAN
  • idea of "computational appendix" or "computational letters" as low-cost e-jounral.
  • discussion of IPython and IJulia as tools to do this.

• TOPIC: How to make toolkit universally available? [MC: Sylvain / Spencer / Chase / Pablo]

  • We would like the tools to be quickly usable by anyone. One suggestion is to build a Python stack organized with conda, and explicitly outline the requirements.
  • Another approach is to build out a virtual machine a la Michael Creel
    • One option here would be to create a docker image for local installation as well as aws/gcloud images for cloud-based execution (can also be built on docker images).

• TOPIC: What about data?

  • Code for modeling is important, but so is code for data
  • Could IMF contribute some API to a public subset of their data (more a question for the IMF guys, not really an item for discussion here)? World bank is a good example of success here
  • We could provide convenient/idiomatic (i.e. lean on pandas in python, DataFrames.jl in Julia) interfaces to this API in chosen languages.
  • Provide similar API for other data sources (FRED, World-Bank, Fama-French, ect.) -- mostly just very small wrappers around existing code to provide unified interface across data sources

• TOPIC: Licenses for code? [Bill]

  • We don't want to discourage usage.

• Topic: Events to foster participation? [Pablo]

  • Hackathons, Google Summer of Code, "EconExchange"
  • [Pablo and / or Bill]

• Topic: Stackoverflow for Economists? [Nathan]

  • "Conspiring" for success of the official SE economics
  • Private options if official one fails
  • Private options for govt / researchers

• Nature of next steps [Nathan]

  • eg. connect with SCE