Retracted articles use less free and open-source software and cite it worse

Abstract

Abstract As an essential mechanism of scientific self-correction, articles are retracted for many reasons, including errors in processing data and computation of results. In today’s data-driven science, the validity of research data and results significantly depends on the software employed. We investigate the relationship between software usage and research validity, eventually leading to article retraction, by analyzing software mentioned across 1,924 retraction notices and 3,271 retracted articles. We systematically compare software mentions and related information with control articles sampled by coarsened exact matching by recognizing publication year, scientific domain, and journal rank. We identify article retractions caused by software errors or misuse and find that retracted articles use less free and open-source software, hampering reproducible research and quality control. Moreover, such differences are also present concerning software citation, where retracted articles less frequently follow software citation guidelines regarding free and open-source software.