The problem
Source: 'Inside the black box' report
- Black boxes obfuscate methods, reduce trust in research, stifle innovation and reduce the ability of future work to build on (and cite) your research
The problem
Source: 'Inside the black box' report
Black boxes obfuscate methods, reduce trust in research, stifle innovation and reduce the ability of future work to build on (and cite) your research
Academic research was developed at a time when computers, let alone open source software and reproducible code, were available: black boxes are the norm
The solution: get your code out there
Source:
Lovelace, Robin 2021: Open Source Tools for Geographic Analysis in Transport Planning. Journal of Geographical Systems. https://doi.org/10.1007/s10109-020-00342-2, accessed January 17, 2021.
Preparing code for publication
At a minum
- Code hosting website (e.g. GitHub)
- Good README with instructions to run code (+optional badges)
- Minimum example (synthetic?) input dataset
- Clear directory structure
- Select and follow a style guide
- Small readable 'chunks' (functions/scripts)
Advanced
- Active issue tracker
- Packaging + documentation
- Code review
- Continuous integration
- Community chat (e.g. discord)
Example: stplanr
Demo of packaged code stplanr (Lovelace and Ellison 2018) https://docs.ropensci.org/stplanr/
Example of code for a paper: https://github.com/Robinlovelace/odjitter
Beyond code 2: open access tools
- Outlines 'network effects' of open research and putting things 'out there'
- Conclusion: open access is particularly important for policy relevant research
Case study of publishing code
Code processing confidential data hosted online: https://github.com/npct/pct-scripts/blob/master/03.2_school_prepare_OD_file.R
Source: Goodman, et al. 2019. Scenarios of Cycling to School in England, and Associated Health and Carbon Impacts: Application of the ‘Propensity to Cycle Tool.’ Journal of Transport & Health 12: 263–278.
Aggregate outputs published
- We commissioned a dataset from DfE, sensitive data, processed securely
- Open code and (more importantly) aggregated derived data increased impact
Beyond code 3: Community engagement
Sharing code happens in context of open source communities
They are usually friendly communities
Getting involved can lead to collaborations
milan_car_crashes <- data.frame( ID = 1:5, # Reproducible example: X = c(1513037, 1513008, 1515473, 1514039, 1515748), Y = c(5034945, 5034750, 5036177, 5036820, 5037396))
Resulting Publication: Gilardi, A., Mateu, J., Borgoni, R., Lovelace, R., 2022. Multivariate hierarchical analysis of car crashes.
Thanks for listening!
Further information: robinlovelace.net + @robinlovelace on Twitter
see slides infoRming policy talk for R-focussed take.
