Posts about reproducibility guidelines

The reproducibility challenge – what researchers need

Within the Open Science discussions, the current call for “reproducibility” comes from the raising awareness that results as presented in research papers are not as easily reproducible as expected, or even contradicted those original results in some reproduction efforts. In this context, transparency and openness are seen as key components to facilitate good scientific practices, as well as scientific discovery. As a result, many funding agencies now require the deposit of research data sets, institutions improve the training on the application of statistical methods, and journals begin to mandate a high level of detail on the methods and materials used. How can researchers be supported and encouraged to provide that level of transparency? An important component is the underlying research data, which is currently often only partly available within the article. At Elsevier we have therefore been working on journal data guidelines which clearly explain to researchers when and how they are expected to make their research data available. Simultaneously, we have also developed the corresponding infrastructure to make it as easy as possible for researchers to share their data in a way that is appropriate in their field. To ensure researchers get credit for the work they do on managing and sharing data, all our journals support data citation in line with the FORCE11 data citation principles – a key step in the direction of ensuring that we address the lack of credits and incentives which emerged from the Open Data analysis (Open Data - the Researcher Perspective https://www.elsevier.com/about/open-science/research-data/open-data-report ) recently carried out by Elsevier together with CWTS. Finally, the presentation will also touch upon a number of initiatives to ensure the reproducibility of software, protocols and methods. With STAR methods, for instance, methods are submitted in a Structured, Transparent, Accessible Reporting format; this approach promotes rigor and robustness, and makes reporting easier for the author and replication easier for the reader.

Code and Data for the Social Sciences: A Practitioner’s Guide

This handbook is about translating insights from experts in code and data into practical terms for empirical social scientists. We are not ourselves software engineers, database managers, or computer scientists, and we don’t presume to contribute anything to those disciplines. If this handbook accomplishes something, we hope it will be to help other social scientists realize that there are better ways to work. Much of the time, when you are solving problems with code and data, you are solving problems that have been solved before, better, and on a larger scale. Recognizing that will let you spend less time wrestling with your RA’s messy code, and more time on the research problems that got you interested in the first place.

Developing Standards for Data Citation and Attribution for Reproducible Research in Linguistics

While linguists have always relied on language data, they have not always facilitated access to those data. Linguistic publications typically include short excerpts from data sets, ordinarily consisting of fewer than five words, and often without citation. Where citations are provided, the connection to the data set is usually only vaguely identified. An excerpt might be given a citation which refers to the name of the text from which it was extracted, but in practice the reader has no way to access that text. That is, in spite of the potential generated by recent shifts in the field, a great deal of linguistic research created today is not reproducible, either in principle or in practice. The workshops and panel presentation will facilitate development of standards for the curation and citation of linguistics data that are responsive to these changing conditions and shift the field of linguistics toward a more scientific, data-driven model which results in reproducible research.

Open for Comments: Linguistics Data Interest Group Charter Statement

Data are fundamental to the field of linguistics. Examples drawn from natural languages provide a foundation for claims about the nature of human language, and validation of these linguistic claims relies crucially on these supporting data. Yet, while linguists have always relied on language data, they have not always facilitated access to those data. Publications typically include only short excerpts from data sets, and where citations are provided, the connections to the data sets are usually only vaguely identified. At the same time, the field of linguistics has generally viewed the value of data without accompanying analysis with some degree of skepticism, and thus linguists have murky benchmarks for evaluating the creation, curation, and sharing of data sets in hiring, tenure and promotion decisions.This disconnect between linguistics publications and their supporting data results in much linguistic research being unreproducible, either in principle or in practice. Without reproducibility, linguistic claims cannot be readily validated or tested, rendering their scientific value moot. In order to facilitate the development of reproducible research in linguistics, The Linguistics Data Interest Group plans to develop the discipline-wide adoption of common standards for data citation and attribution. In our parlance citation refers to the practice of identifying the source of linguistic data, and attribution refers to mechanisms for assessing the intellectual and academic value of data citations.

Reproducible Data Analysis in Jupyter

Jupyter notebooks provide a useful environment for interactive exploration of data. A common question I get, though, is how you can progress from this nonlinear, interactive, trial-and-error style of exploration to a more linear and reproducible analysis based on organized, packaged, and tested code. This series of videos presents a case study in how I personally approach reproducible data analysis within the Jupyter notebook.