Goals for Staff Biostatisticians
This is for submission to VUMC HR beginning November, 2015.
Goal 1: Practice Reproducible Research
Routinely use and continually improve upon good practices that ensure your work is trustworthy, transparent, and reproducible.
Expectations / rubrics:
- Goal met at 0%: Some analyses are run using GUI interfaces, scripts and data are saved in haphazard locations, and/or analyses are performed in a manner that is difficult to replicate a year after the analysis was last run.
- 50%: All analyses are run with a re-executable code and saved in a logical, easy to find structure. Analyses are able to be reliably rerun a year after the analysis was last run.
- 75%: The majority of major analyses are done within reproducible reports that contain important interpretation and documentation within the report.
- 100%: All major analyses are done with reproducible reports where the de-identified (free of protected health information) reports and code that generated them are publicly shared.
Goal 2: Create Reusable Code
Write your code to be easily adapted to multiple projects. As appropriate for your work, modularize and functionize to achieve greater long-term efficiency.
Expectations / rubrics:
- Goal met at 0%: All code is project specific, such that a new project requiring similar tasks receives no benefit from the previous code. Code contains PHI (protected health information).
- 50%: Code is modularized and somewhat adaptable, but not well functionized so that extensive revisions are required for each project.
- 75%: Code is often functionized to allow specific tasks need for projects to be completed quickly and reliably. Any task that has been encountered at least three times has an attempted functionalized solution.
- 100%: Code is highly functionized to allow new projects with similar needs to be quickly and reliably completed. Functions are well documented and shared publicly, e.g. via a GitHub repository.