Functional R code is rare in species distribution and abundance papers

Abstract

Analytic reproducibility is important for scientific credibility in ecology, but the extent to which scientific literature meets this criterion is not well understood. We surveyed 497 papers published in 2018–2022 in 9 ecology‐related journals. We focused on papers that used hierarchical models to estimate species distribution and abundance. We determined if papers achieved two components of analytic reproducibility: (1) availability of data and code, and (2) code functionality. We found that 28% of papers made data and code available, and 7% of papers provided code that ran without errors. Our findings indicate that analytic reproducibility remains the exception rather than the rule in ecology literature. We recommend authors (1) test code in a separate clean environment; (2) simplify code structure; (3) minimize software packages used; and (4) minimize code run time. We suggest journals (1) validate authors' provided open data statements and URLs; (2) recommend that code and data be shared in a separate repository rather than as appendices; and (3) elevate the status of code and data during review. We suggest these guidelines can aid the ecology community by improving the scientific reproducibility and credibility of ecological research.