Science is allegedly in the midst of a reproducibility crisis, but questions of reproducibility and related principles date back nearly 80 years. Numerous controversies have arisen, especially since 2010, in a wide array of disciplines that stem from the failure to reproduce studies or their findings:biology, biomedical and preclinical research, business and organizational studies, computational sciences, drug discovery, economics, education, epidemiology and statistics, genetics, immunology, policy research, political science, psychology, and sociology.
This monograph defines terms and constructs related to reproducible research, weighs key considerations and challenges in reproducing or replicating studies, and discusses transparency in publications that can support reproducible research goals. It attempts to clarify reproducible research, with its attendant (and confusing or even conflicting) lexicon and aims to provide useful background, definitions, and practical guidance for all readers.
Among its conclusions: First, researchers must become better educated about these issues, particularly the differences between the concepts and terms. The main benefit is being able to communicate clearly within their own fields and, more importantly, across multiple disciplines. In addition, scientists need to embrace these concepts as part of their responsibilities as good stewards of research funding and as providers of credible information for policy decision making across many areas of public concern. Finally, although focusing on transparency and documentation is essential, ultimately the goal is achieving the most rigorous, high-quality science possible given limitations on time, funding, or other resources.
Reproducible research in computational sciences: A use case for uncertainty quantification using Jupyter Notebooks