Goal

The goal of this workshop is to make early-career researchers in the social sciences (e.g. Ph.D. students and early faculty) aware of computational tools and toolkits that will allow them to more efficiently and reproducibly conduct their research. We will showcase computer-oriented techniques and tools, from basic command line tools on Windows, Linux, and Mac, to version control, to optimization and parallelization techniques. These methods are useful not just for high-performance computing, but by necessity make research reproducible. If you touch a computer as part of your research, you should come!

More goals and non-goals

This workshop is designed to open your eyes to the possibilities, scratching the surface, but mostly not diving into any particular depths. The goal is NOT to teach a full course on SAS, Stata, Matlab, R, Python, MPI, Fortran, etc. - there are other classes for that. We will teach just enough of each programming language to be able to highlight additional techniques. There will be hands-on training on a few systems (TBD).

We will discuss ways to structure your computional projects with an eye on efficiency during development and reproducibility after publication (by AEA standards). Additional topics include reproducibility when data may be private, computation when compute resources are scarce (parallel computing, cloud computing), automation in the service of reliability and reproducibility, and other topics. We will provide hands-on training in at least one advanced computing technique.

You and the workshop

We highlight that this is a workshop - we will work on problems as a group, drawing on expertise in the “room” as needed. If you have a specific question, and want to work on it, we may do so. If you want to primarily listen, that’s fine too. You should expect to do some exercises each day. We have a two-day break (to not step on the Economics Department’s TA training), and reconvene on the Friday (Aug 19) to discuss outcomes, questions, and next steps for some.

Target group

Second year Ph.D. and higher, and faculty, in Economics or other social sciences. If you haven’t participated in the workshop in the past, or want a refresher, you should participate! If you want to chat about the computational aspects of social sciences, or reproducibility, come and join the workshop!

Tentative schedule

Details coming soon:

Monday

Hands-on creation of a reproducible project, ready to be uploaded to a journal replication archive.

Computational empathy: why you should care
How to structure a project
- Some basics (TIER Protocol, confidential data)
Git, Version control
- Naming of programs: incomplete examples
- Framework: Carpentries’ “Git basics”
Why you should learn (and love) the command line
- Framework: Carpentries’ “The Unix Shell”
- Some brief examples
Documenting it all: Readme
- Template README by social science data editors (Walkthrough)
- Data citations, and Tricky data citations
Data management
- Example: API use and local temporary storage, see labordynamicsinstitute/alfred_example
- Notes from live presentation: not this year…
Sundry
- Reproducibility and Ethics/ sharing data
- Computing at Cornell: BioHPC but more on Friday.

Tuesday

Follow-up to Monday, plus some high-performance computing tricks, and how to deposit your own data at a trusted repository (and what all of that means)

Computing environments
- Yours, mine, theirs: why how you run Stata, or R, or Python matters
- Some solutions (habits, instructions, Docker)
- Installing packages, and how not to install them
Docker for reproducible research, and why you should know about it, even if you don’t use it
- Running R in the cloud
- Docker for Stata
  - Docker image: https://hub.docker.com/r/dataeditors/stata17/
  - How Stata in the cloud works: template with instructions
- Using Github Codespaces
- Using Stata on Github Codespaces
- (note: to make it work, you need to be a member of a Github “organization”)
Parallel computing
- Basics in R (see tutorial)
  - mclapply (done)
  - Creating a cluster (not done)
- Package in Stata (not done in the workshop)
- OS job schedulers (e.g., SLURM at BioHPC and on ECCO)
Data management in a parallel computing world
- Time cost
- File systems
- Database systems (MariaDB, Neo4j, etc.)
Creating an archive of data/code/replication packages
- Archives/repositories:
  - openICPSR for AEA
  - Harvard Dataverse
  - Zenodo for AEA
  - Zenodo Sandbox
  - OSF (not a full archive)
- An example of an archive ready workflow, from A (acquisition) to Z (Zenodo) and beyond (overall tutorial, and the key slides) (not done)
- Zenodo upload script
  - Resulting “archive” on Sandbox
- Dataverse upload Github Action (not done)

Friday

Details on computing infrastructure for social scientists (in particular economists) at Cornell and elsewhere. When to use what.

Computational Tools for Social Scientists Workshop