Skip to content
Shauna edited this page Oct 21, 2019 · 13 revisions

See past roadmaps for details on past roadmaps.

3 Month Roadmap

End Date: January 15th 2020

Ordered By Goal

Integrating Econ-Ark and Dolo

  • arrange further meeting w/ Sebastian, Pablo, Chris to explain Dolo [SB]
  • implement three models (Krusell-Smith, Ayiagari, CSTWMPC) using hybrid of Hark and Dolark, including formal mathematical description of the models being solved [SB]

Improve Testing

  • get notebooks integrated into testing [MS]
  • full coverage of core modules [MS; blocked by CDC who needs to confirm which modules to cover]

Improve Documentation/Information

  • re-organize project repositories [SGM via Sam Brice]
  • update workflow for notebooks that are documentation vs not documentation [SGM]
  • create comprehensive index of content in all of the repos and notebooks [SGM via Andrij Stachurski]
  • synthesize existing “how to use HARK” materials (or create new ones as needed) into walkthrough(s) using Jupyter notebook, based on content from the HARK Manual [SB]
  • speed up loading of REMARKs [MS]

Misc

  • stop supporting Python 2 [MS]
  • determine a plan for linking HARK versions to notebooks [SGM; MS can help]
  • keep an eye out for low-effort ways to continue supporting/building community [everyone]
  • PKM needs to finish work re: risky investments and housing choice [SGM]
  • respond to SFF, possibly with arkhiver project [CDC]

Ordered By Contributor

CDC

  • respond to SFF, possibly with arkhiver project [CDC]
  • full coverage of core modules [MS; blocked by CDC who needs to confirm which modules to cover]

SGM

  • determine a plan for linking HARK versions to notebooks [SGM; MS can help]
  • PKM needs to finish work re: risky investments and housing choice [SGM]
  • re-organize project repositories [SGM via Sam Brice]
  • update workflow for notebooks that are documentation vs not documentation [SGM]
  • create comprehensive index of content in all of the repos and notebooks [SGM via Andrij Stachurski]

SB

  • arrange further meeting w/ Sebastian, Pablo, Chris to explain Dolo [SB]
  • implement three models (Krusell-Smith, Ayiagari, CSTWMPC) using hybrid of Hark and Dolark, including formal mathematical description of the models being solved [SB]
  • synthesize existing “how to use HARK” materials (or create new ones as needed) into walkthrough(s) using Jupyter notebook, based on content from the HARK Manual [SB]

MS

  • get notebooks integrated into testing [MS]
  • full coverage of core modules [MS; blocked by CDC who needs to confirm which modules to cover]
  • speed up loading of REMARKs [MS]
  • stop supporting Python 2 [MS]

12 Month Roadmap

End Date: January 15th 2020

Misc

  • End Python 2 support

Refactor HARK

  • Refactor deep model structure
  • Document new deep model structure

Begin planning HARK 1.0

  • Generate a design document for HARK 1.0. What is the overarching structure, and what other frameworks would fit within it?
    • Foundation for modeling improvements is Pablo's work on dolo/dolARK
    • Map out future "frameworks" besides HARK
      • Incorporate Jackie’s work with Mesa and agent-based modeling, and/or

More Distant Goals

The REMARK repo is a start at defining a standard for reproducible results in computational economics, but only a start. Eventually, we would like to have a system set up so that

  1. People can submit a request for a project to be turned into a reproducible archive, and there will be an automated system that checks their submission to make sure that it satisfies a set of baseline criteria
    • The first version of these criteria will be identical to the criteria for the projected journal we are thinking of creating
    • Since that journal will start off as a clone of the Journal of Open Source Software, we will adopt all of the requirements that are already in place for JOSS submissions
    • In addition to those requirements, we will need a system for creating a standalone archive that in principle anybody with a functioning computer could use to reproduce the results
  2. The right technology for doing this is probably a Docker container
  3. There are lots of thorny issues we will have to tackle if we want to make a system that can handle proprietary code and data.
    • We will put off those issues for now; the first version of the system should be designed to handle only open source projects using tools and languages with modern version control technologies (like python and julia)