Code Review
Some useful text here to explain who and what and when this should be done.. TBC
- Basic checks:
- Code is on github
- Study is organised as an R project
- Renv is used to list all dependencies needed
- Study code has a clear logical flow, with any particularly long scripts split up into separate files
- Study code doesn’t have a lot of complex, custom code (that should be in a package with tests)
- The code runs on a 100k dataset without error
- How are the results visualised and reported?
- Is there a shiny to go with the study code?
- Review results for plausibility
- Connection details are not displayed in scripts such as CodeToRun
- Check whether the code does what is intended:
- Does the code match the protocol?
- Have any analyses been missed?
- For each analysis, are cohorts defined in the right way (e.g. typically no exclusion criteria for an incidence outcome) - this has been the most common source of issues
- Check whether the code can be optimised:
- Is any code repeated unnecessarily?
- Can code be simplified?
- Review the sql that gets executed for any obvious inefficiencies