Funders of scientific research are well-positioned to guide scientific discoveries by enabling and incentivizing the most rigorous and transparent methods. This resource hub provides examples of best practices currently employed by funders of biomedical, educational, and social sciences. These recommendations and templates provide useful tools so that you the funder can learn from others how to best shift norms in the entire research community.
Transparency into the items generated over the course of a study (e.g. the data, questionnaires, stimuli, physical reagents, or analytical code) benefit the entire research community. Consumers of knowledge have more trust in the underlying findings, researchers have more materials to reuse, and the generators of these items have a well curated set of materials to use in future years. Curating these items after the study is completed can be a time-consuming task, so we encourage the use of tools that combine project management and preservation into the same platform, which underlies the design and development of the OSF.
Citation of articles is routine. Similar expectations can and should be applied to citation of data, code, and materials to recognize and credit these as original intellectual contributions. Funders can encourage this practice (and by extension incentivize the sharing of these items) by clarifying standards for data, materials, and code citations in applications, grant reports, and papers reporting the results of funded research.
Reporting Guidelines and Checklists for Design and Analysis Transparency
Standards for reporting research designs and analyses should maximize transparency about the research process and minimize potential for vague or incomplete reporting of the methodology. Such standards are highly discipline-specific, and recommendations can provide guidance on how to identify and use relevant checklists. The use of these checklists should seek to maximize:
Researchers are highly motivated to discover unexpected relationships or effects, but are also looking to ensure that these findings are robust. Unfortunately, the tension between these two motivations lead to (mis)application of statistical tools, whereby trends in a dataset are used to generate a hypothesis, which is then tested by using the same dataset that was used to create it. This and other questionable practices, such as repeatedly trying slightly different analyses, invalidates the most common statistical tools and corrodes trust in scientific findings.
Preregistration makes a clear distinction between hypothesis-testing (confirmation) and hypothesis-generation (discovery or exploration).
Preregistration is especially important if you want to support research projects whose purpose is to make an inferential claim from a sampled population to some wider population. This can apply to casual inference studies such as randomized, controlled trials (RCTs), but is also relevant to other study designs. Observational studies, epidemiological research, or other methods that assume any a-priori hypothesis can all benefit from preregistration.
Preregistration may not be needed in purely exploratory research, theory/model development, or purely descriptive research. In these cases, the recommended best practice is transparently documenting workflows for preservation and reuse. This can be done with version controlled repositories such as GitHub, general purpose project tools such as OSF, or any electronic lab notebook (ELN).
"The number NHLBI trials reporting positive results declined after the year 2000. Prospective declaration of outcomes in RCTs, and the adoption of transparent reporting standards, as required by clinicaltrials.gov, may have contributed to the trend toward null findings."
"Signs of bias from lack of trial protocol registration were found with non-registered trials reporting more beneficial intervention effects than registered ones."
Papageorgiou, S. N., Xavier, G. M., Cobourne, M. T., & Eliades, T. (2018). Registered trials report less beneficial treatment effects than unregistered ones: A meta-epidemiological study in orthodontics. Journal of Clinical Epidemiology. https://doi.org/10/gdzqnb
"Strong results are 40 percentage points more likely to be published than are null results and 60 percentage points more likely to be written up. We provide direct evidence of publication bias and identify the stage of research production at which publication bias occurs: Authors do not write up and submit null findings."
See a curated list of studies on the need for and benefits of other open practices here.
Registered Reports build on the benefits of preregistration by ensuring that preregistered research will be published regardless of outcome. In order to reach that threshold, the preregistered proposal is first peer reviewed by the journal to evaluate the importance of the research questions and the ability of the proposed study to address them. High quality studies are then granted "in-principle acceptance" (IPA) and published whether results are significant or null.
Funders can support this process by creating incentives for researchers to use it or by partnering with journals to combine grant review with publication review, thus creating a more efficient and less biased research system.
A replication is a study that seeks to recreate a previously-published experiment. For replications, any outcome could be considered evidence that may increase or decrease confidence about a claim from previously-conducted research. Replication is perceived as boring, uncreative, confirmatory work of science. This misperception leads to minimal funding or publishing replication studies, and contributes to the Replication Crisis by devaluing the importance of replications.
No single study, whether novel or replication, can provide a definitive answer to any claim. Outcomes from replication studies can add to the available information supporting a claim, which can lead to refinement of theory to the generation of new, testable predictions. In this light, replications are central to bolstering the evidence for/against a specific claim as well as the discovery of new, interesting phenomena.
The outdated image of science being conducted by a lone genius still affects how scientists are perceived by the public today. Awards that highlight a single individual's achievement are one example of how this notion persists. Unfortunately, this image is at odds with how science is largely practiced. Funders can support a more collaborative community that recognizes this reality and supports better research.