Routine data sharing, defined here as the publication of the primary data and any supporting materials required to interpret the data acquired as part of a research study, is still in its infancy in psychology, as in many domains. Nevertheless, with increased scrutiny on reproducibility and more funder mandates requiring sharing of data, the issues surrounding data sharing are moving beyond whether data sharing is a benefit or a bane to science, to what data should be shared and how. Here, we present an overview of these issues, specifically focusing on the sharing of so-called "long tail" data, that is, data generated by individual laboratories as part of largely hypothesis-driven research. We draw on experiences in other domains to discuss attitudes toward data sharing, cost-benefits, best practices and infrastructure. We argue that the publishing of data sets is an integral component of 21st-century scholarship. Moreover, although not all issues around how and what to share have been resolved, a consensus on principles and best practices for effective data sharing and the infrastructure for sharing many types of data are largely in place. (PsycINFO Database Record

Modern research often involves the collection or analysis of data and the use of specialized computer algorithms. Traditional text articles thus provide only partial documentation of a research study. Readers have limited ability to reproduce or utilize work if the source data are not available or if it relies on an algorithm that is described, but code is not provided. Fortunately, a wide variety of tools are now available to support the publication of research data and code. The effort required to publish data is now relatively small, and the benefits can be immense. This opinion article discusses trends toward increased sharing in academic publishing. It describes opportunities and resources to support data and code sharing and describes the benefits for both authors and readers. Finally, it discusses how Earthquake Spectra is providing resources and enhancing its policies to establish the sharing of data as the default procedure when publishing in the journal, and encourage the sharing of code and other resources.

Sharing of code supports reproducible research but fewer journals have policies on code sharing compared to data sharing, and there is little evidence on researchers’ attitudes and experiences with code sharing. Before introducing a stronger policy on sharing of code, the Editors and publisher of the journal PLOS Computational Biology wished to test, via an online survey, the suitability of a proposed mandatory code sharing policy with its community of authors. Previous research has established, in 2019, 41% of papers in the journal linked to shared code. We also wanted to understand the potential impact of the proposed policy on authors' submissions to the journal, and their concerns about code sharing.We received 214 completed survey responses, all of whom had generated code in their research previously. 80% had published in PLOS Computational Biology and 88% of whom were based in Europe or North America. Overall, respondents reported they were more likely to submit to the journal if it had a mandatory code sharing policy and US researchers were more positive than the average for all respondents. Researchers whose main discipline is Medicine and Health sciences viewed the proposed policy less favourably, as did the most senior researchers (those with more than 100 publications) compared to early and mid-career researchers.The authors surveyed report that, on average, 71% of their research articles have associated code, and that for the average author, code has not been shared for 32% of these papers. The most common reasons for not sharing code previously are practical issues, which are unlikely to prevent compliance with the policy. A lack of time to share code was the most common reason. 22% of respondents who had not shared their code in the past cited intellectual property (IP) concerns - a concern that might prevent public sharing of code under a mandatory code sharing policy. The results also imply that 18% of the respondents’ previous publications did not have the associated code shared and IP concerns were not cited, suggesting more papers in the journal could share code.To remain inclusive of all researchers in the community, the policy was designed to allow researchers who can demonstrate they are legally restricted from sharing their code to be granted an exemption to public sharing of code.As a secondary goal of the survey we wanted to determine if researchers have unmet needs in their ability to share their own code, and to access other researchers' code. Consistent with our previous research on data sharing, we found potential opportunities for new products or features that support code accessibility or reuse. We found researchers were on average satisfied with their ability to share their own code, suggesting that offering new products or features to support sharing in the absence of a stronger policy would not increase the availability of code with the journal's publications.

Do Journal Code-sharing Policies Increase Code Availability? | Zenodo Skip to main You are using an outdated browser. Please upgrade your browser to improve your experience. Open Science Conference 2023 Published July 24, 2023 | Version v1 Presentation Open Do Journal Code-sharing Policies Increase Code Availability? Authors/Creators Aya Bezine 1 Alfredo Sánchez-Tójar 2 Antica Culina 3 Marija Purgar 3 Show affiliations 1. Bielefeld University 2. Bielefeld University; 3. Ruder Boskovic Institute Description Ensuring reproducibility is one of the main goals of open science. To achieve reproducibility of scientific results, data and analytical computer code (if used) should ideally be openly available. Thus far, efforts have mostly focused on making research data open and FAIR (Findable, Accessible, Interoperable, and Reusable), whereas code-sharing has only recently started to gain attention. A recent review of the state of code availability in ecological journals with code-sharing policies showed that, despite the policies, code-sharing is alarmingly low, suggesting that those policies are not adhered to by most authors. This project builds on that recent review to explicitly test whether journal code-sharing policies increase code availability. For that, we compare code availability between 14 journals with (n = 346 articles) and 13 journals without code-sharing policies (n = 350 articles) for the period between 2015 and 2019. Our study aims to provide essential information to help improve code-sharing and journal code-sharing policies in ecology and other fields, with the ultimate goal of making science more reproducible in the short- and long-term. Our findings thus far provide evidence that the implementation of a journal code-sharing policy does not result in a substantial proportion of code-sharing. However, it does appear to contribute to an overall increase in the availability of code, which is a positive outcome. When considering eligible articles published, the percentage of papers that shared their code was significantly higher in journals with a code-sharing policy (27%) compared to journals without such a policy (3%). Another noteworthy observation is the progressive increase in code-sharing within journals lacking a code-sharing policy over time. Specifically, only 1% of papers published between 2015-2016 shared their code, whereas this figure rose to 4% between 2018-2019. In sum, our results suggest that although code-sharing policies lead to an increase in code availability, the scientific community still needs to ensure that those policies are truly implemented. For that, we suggest that journals must provide better support for researchers and have a system for ensuring that policies are adhered to. Files OSC2023_Bezine.pdf Files (1.2 MB) Name Size Download all OSC2023_Bezine.pdf md5:282eb3d43bc15d3618ecced5cd78d9d3 1.2 MB Preview Download 202 Views 75 Downloads Show more details All versions This version Views Total views 202 202 Downloads Total downloads 75 75 Data volume Total data volume 103.2 MB 103.2 MB More info on how stats are collected....

tly reproduced (Supplement 1 lines 226-263). Deviations from the study protocol. Any deviations from the analysis plan, eg, data transformations, should be transparently reported (Supplement 1 lines 348-376). If any analyses are performed, but not reported in the published research article, a rationale for not reporting the findings should be provided in the code. Recommendation 5: focus on accessibility of code and data Last, efforts taken to improve transparency or comprehensibility of code are fruitless if the analytical code and data are not made available. 1 It may not always be possible to share analytical data, and when individual-level participant data is shared, researchers should be mindful that this is in compliance with participant consent and local data protection laws. They should take precautions to protect privacy and prevent re-identification, for example by masking quasi-identifiers and applying k-anonymity. However, even when data cannot be shared freely, sharing the corresponding code is still informative to get an overview of data management and the performed statistical analyses. 25 For instance, providing an overview of the data cleaning process and the handling of missing data can help the reader to better understand the study, even without accessing the data. Moreover, researchers can consider sharing metadata such as variable types and labels, which facilitates reproducibility for future researchers working with the same dataset (eg, when working with the UK Biobank). 4 This approach is in line with the FAIR guidelines proposing that data should be findable, accessible, interoperable and reusable. 26 If data cannot be shared freely, authors can still state how the data can be accessed if permission is obtained. 1 We identified and listed different options to share data and analytical code, each with their own advantages and disadvantages in table 1 . Sharing code and data as a supplement to a paper is the most direct way of improving the re

Online product reviews underpin nearly all e-shopping activities. The high volume of data, as well as various online review quality, puts growing pressure on automated approaches for informative content prioritization. Despite a substantial body of literature on review helpfulness prediction, the rationale behind specific feature selection is largely under-studied. Also, the current works tend to concentrate on domain- and/or platform-dependent feature curation, lacking wider generalization. Moreover, the issue of result comparability and reproducibility occurs due to frequent data and source code unavailability. This study addresses the gaps through the most comprehensive feature identification, evaluation, and selection. To this end, the 30 most frequently used content-based features are first identified from 149 relevant research papers and grouped into five coherent categories. The features are then selected to perform helpfulness prediction on six domains of the largest publicly available Amazon 5-core dataset. Three scenarios for feature selection are considered: (i) individual features, (ii) features within each category, and (iii) all features. Empirical results demonstrate that semantics plays a dominant role in predicting informative reviews, followed by sentiment, and other features. Finally, feature combination patterns and selection guidelines across domains are summarized to enhance customer experience in today's prevalent e-commerce environment. The computation PLoS One PLoS ONE 440 plosone 101285081 plos PLoS ONE 1932-6203 PLOS PMC6927604 PMC6927604.1 6927604 6927604 31869404 10.1371/journal.pone.0226902 PONE-D-19-13368 1 Research Article Social Sciences Linguistics Grammar Syntax Social Sciences Linguistics Lexicons Social Sciences Linguistics Semantics Research and Analysis Methods Research Assessment Reproducibility Social Sciences Linguistics Grammar Phonology Vocabulary Social Sciences Linguistics Linguistic Morphology Computer and Information Sciences Data Management Metadata Social Sciences Linguistics Grammar Feature selection for helpfulness prediction of online product reviews: An empirical study Feature selection for review helpfulness prediction http://orcid.org/0000-0001-9676-1186 Du Jiahua Conceptualization Data curation Investigation Methodology Software Validation Visualization Writing – original draft 1 Rong Jia Methodology Supervision Validation Writing – review & editing 1 2 * Michalska Sandra Validation Visualization Writing – review & editing 1 Wang Hua Supervision 1 Zhang Yanchun Supervision 1 1 Institute of Sustainable Industries & Liveable Cities, Victoria University, Melbourne, VIC, Australia 2 Faculty of Information Technology, Monash University, Clayton, VIC, Australia Khan Farhan Hassan Editor College of EME, NUST, PAKISTAN Competing Interests: The authors have declared that no competing interests exist. Despite a substantial body of literature on review helpfulness prediction, the rationale behind specific feature selection is largely under-studied. Also, the current works tend to concentrate on domain- and/or platform- dependent feature curation, lacking wider generalization. Moreover, the issue of result comparability and reproducibility occurs due to frequent data and source code unavailability. This study addresses the gaps through the most comprehensive feature identification, evaluation, and selection. To this end, the 30 most frequently used content-based features are first identified from 149 relevant research papers and grouped into five coherent categories. pmc-status-qastatus 0 pmc-status-live yes pmc-status-embargo no pmc-status-released yes pmc-prop-open-access yes pmc-prop-olf no pmc-prop-manuscript no pmc-prop-legally-suppressed no pmc-prop-has-pdf yes pmc-prop-has-supplement no pmc-prop-pdf-only no pmc-prop-suppress-copyright no pmc-prop-is-real-version no pmc-prop-is-scanned-article no pmc-prop-preprint no pmc-prop-in-epmc yes pmc-license-ref CC BY Data Availability This paper used the Amazon 5-core dataset which is publicly available on http://jmcauley.ucsd.edu/data/amazon/ . The configuration of training-validation-testing splits and source code for our experiments containing pre-processing, feature extraction, and forward feature selection, are provided with open access on https://github.com/tokawah/Helpfulness-Feature-Selection . Data Availability This paper used the Amazon 5-core dataset which is publicly available on http://jmcauley.ucsd.edu/data/amazon/ . The configuration of training-validation-testing splits and source code for our experiments containing pre-processing, feature extraction, and forward feature selection, are provided with open access on https://github.com/tokawah/Helpfulness-Feature-Selection . Instead of evaluating the entire feature set, the study allows for performance-oriented feature selection under multiple scenarios. Such flexibility can effectively justify (not) selecting certain features. As a result, feature combination patterns and selection guidelines across domains are summarized, offering valuable insights into general feature selection for helpfulness prediction. The publicly available source code and datasets ensure result comparability and reproducibility of the study. Fourth, the source code, dataset splits, pre-processed reviews, and extracted features have been released for result reproducibility, benchmark studies, and further improvement. The remaining of the study is organized as The following rules are adopted for feature list compilation: (i) features mentioned at least three times over the entire paper collection to exclude rare features, (ii) removal of human-annotated features due to expensive manual annotation process, and (iii) inclusion of only content-based features to support platform-independent generalizability and transferability. As a results, 27 feature candidates are identified. 10.1371/journal.pone.0226902.t001 Table 1 Features used in the analysis.