Speaker
Description
Reproducibility in ecological research is a known problem with a less-known solution: not one tool, but a small, composable set of them. This poster introduces four R packages that, used together,
address most of the practical pain points researchers encounter between raw data and published results.
Each panel maps to one package. renv handles the environment itself — separate package libraries per project, version numbers recorded, no more "it works on my machine." Crucially, it also means
that a project archived on Zenodo today will still run in five years, after dozens of package updates have broken everything else. pins solves a quieter but equally frustrating problem: shared
datasets that live in too many places, get out of sync, and lack a single source of truth.
testthat is the least obvious inclusion for ecologists, but arguably the most valuable. It lets you write down, in code, what you expect your data to look like — column names, value ranges, absence
of duplicates — and check those expectations automatically at every step of a pipeline. Combined with checkmate for argument validation and testdat for dataframe-specific checks, it turns
assumptions that usually live in a researcher's head into something the computer can verify. Silent data corruption, wrong joins, and unexpected NAs get caught early rather than at submission.
Finally, rmarkdown (and its successor quarto) ties everything together into documents where code, results, and prose coexist and can be re-run at any point.
The central section of the poster walks through a short workflow combining all four: loading a shared species occurrence dataset via pins, running it through a validated cleaning pipeline with
testthat checks at each step, within a renv-locked environment, and producing a reproducible report with rmarkdown.
The goal is practical. Most researchers already use one of these tools. The poster is meant to be the nudge that gets them to use the other three.
| Status Group | Postdoctoral Researcher |
|---|