Explore the vast range of titles available.

Progress in science relies in part on generating hypotheses with existing observations and testing hypotheses with new observations. This distinction between postdiction and prediction is appreciated conceptually but is not respected in practice. Mistaking generation of postdictions with testing of predictions reduces the credibility of research findings. However, ordinary biases in human reasoning, such as hindsight bias, make it hard to avoid this mistake. An effective solution is to define the research questions and analysis plan before observing the research outcomes—a process called preregistration. Preregistration distinguishes analyses and outcomes that result from predictions from those that result from postdictions. A variety of practical strategies are available to make the best possible use of preregistration in circumstances that fall short of the ideal application, such as when the data are preexisting. Services are now available for preregistration across all disciplines, facilitating a rapid increase in the practice. Widespread adoption of preregistration will increase distinctiveness between hypothesis generation and hypothesis testing and will improve the credibility of research findings.

A key component of scientific communication is sufficient information for other researchers in the field to reproduce published findings. For computational and data-enabled research, this has often been interpreted to mean making available the raw data from which results were generated, the computer code that generated the findings, and any additional information needed such as workflows and input parameters. Many journals are revising author guidelines to include data and code availability. This work evaluates the effectiveness of journal policy that requires the data and code necessary for reproducibility be made available postpublication by the authors upon request. We assess the effectiveness of such a policy by (i) requesting data and code from authors and (ii) attempting replication of the published findings. We chose a random sample of 204 scientific papers published in the journal Science after the implementation of their policy in February 2011. We found that we were able to obtain artifacts from 44% of our sample and were able to reproduce the findings for 26%. We find this policy—author remission of data and code postpublication upon request—an improvement over no policy, but currently insufficient for reproducibility.

Publication in peer-reviewed journals is an essential step in the scientific process. However, publication is not simply the reporting of facts arising from a straightforward analysis thereof. Authors have broad latitude when writing their reports and may be tempted to consciously or unconsciously “spin” their study findings. Spin has been defined as a specific intentional or unintentional reporting that fails to faithfully reflect the nature and range of findings and that could affect the impression the results produce in readers. This article, based on a literature review, reports the various practices of spin from misreporting by “beautification” of methods to misreporting by misinterpreting the results. It provides data on the prevalence of some forms of spin in specific fields and the possible effects of some types of spin on readers’ interpretation and research dissemination. We also discuss why researchers would spin their reports and possible ways to avoid it.

Some aspects of science, taken at the broadest level, are universal in empirical research. These include collecting, analyzing, and reporting data. In each of these aspects, errors can and do occur. In this work, we first discuss the importance of focusing on statistical and data errors to continually improve the practice of science. We then describe underlying themes of the types of errors and postulate contributing factors. To do so, we describe a case series of relatively severe data and statistical errors coupled with surveys of some types of errors to better characterize the magnitude, frequency, and trends. Having examined these errors, we then discuss the consequences of specific errors or classes of errors. Finally, given the extracted themes, we discuss methodological, cultural, and systemlevel approaches to reducing the frequency of commonly observed errors. These approaches will plausibly contribute to the self-critical, self-correcting, ever-evolving practice of science, and ultimately to furthering knowledge.

Observational healthcare data, such as electronic health records and administrative claims, offer potential to estimate effects of medical products at scale. Observational studies have often been found to be nonreproducible, however, generating conflicting results even when using the same database to answer the same question. One source of discrepancies is error, both random caused by sampling variability and systematic (for example, because of confounding, selection bias, and measurement error). Only random error is typically quantified but converges to zero as databases become larger, whereas systematic error persists independent from sample size and therefore, increases in relative importance. Negative controls are exposure–outcome pairs, where one believes no causal effect exists; they can be used to detect multiple sources of systematic error, but interpreting their results is not always straightforward. Previously, we have shown that an empirical null distribution can be derived from a sample of negative controls and used to calibrate P values, accounting for both random and systematic error. Here, we extend this work to calibration of confidence intervals (CIs). CIs require positive controls, which we synthesize by modifying negative controls. We show that our CI calibration restores nominal characteristics, such as 95% coverage of the true effect size by the 95% CI. We furthermore show that CI calibration reduces disagreement in replications of two pairs of conflicting observational studies: one related to dabigatran, warfarin, and gastrointestinal bleeding and one related to selective serotonin reuptake inhibitors and upper gastrointestinal bleeding. We recommend CI calibration to improve reproducibility of observational studies.

We describe and demonstrate an empirical strategy useful for discovering and replicating empirical effects in psychological science. The method involves the design of a metastudy, in which many independent experimental variables—that may be moderators of an empirical effect—are indiscriminately randomized. Radical randomization yields rich datasets that can be used to test the robustness of an empirical claim to some of the vagaries and idiosyncrasies of experimental protocols and enhances the generalizability of these claims. The strategy is made feasible by advances in hierarchical Bayesian modeling that allow for the pooling of information across unlike experiments and designs and is proposed here as a gold standard for replication research and exploratory research. The practical feasibility of the strategy is demonstrated with a replication of a study on subliminal priming.

It appears paradoxical that science is producing outstanding new results and theories at a rapid rate at the same time that researchers are identifying serious problems in the practice of science that cause many reports to be irreproducible and invalid. Certainly, the practice of science needs to be improved, and scientists are now pursuing this goal. However, in this perspective, we argue that this seeming paradox is not new, has always been part of the way science works, and likely will remain so. We first introduce the paradox. We then review a wide range of challenges that appear to make scientific success difficult. Next, we describe the factors that make science work—in the past, present, and presumably also in the future. We then suggest that remedies for the present practice of science need to be applied selectively so as not to slow progress and illustrate with a few examples. We conclude with arguments that communication of science needs to emphasize not just problems but the enormous successes and benefits that science has brought and is now bringing to all elements of modern society.

After documenting the existence and exploring some implications of three alternative news narratives about science and its challenges, this essay outlines ways in which those who communicate science can more accurately convey its investigatory process, self-correcting norms, and remedial actions, without in the process legitimizing an unwarranted “science is broken/in crisis” narrative. The three storylines are: (i) quest discovery, which features scientists producing knowledge through an honorable journey; (ii) counterfeit quest discovery, which centers on an individual or group of scientists producing a spurious finding through a dishonorable one; and (iii) a systemic problem structure, which suggests that some of the practices that protect science are broken, or worse, that science is no longer self-correcting or in crisis.

Reproducibility has been one of the major tools science has used to help establish the validity and importance of scientific findings since Philosophical Transactions of the Royal Society was established in 1665. Since that time the process of discovery has evolved to make use of new technologies and methods in a changing regulatory and social environment. The Sackler Colloquium \"Reproducibility of Research: Issues and Proposed Remedies,\" which took place on Mar 8-10, 2017, convened a wide range of research community stakeholders to address our understanding of transparency and reproducibility in the modern research context. Here, Allison et al approach the topic of reproducibility with sensitivity to its complexity, spanning a wide range of issues from data collection and reporting to communication of scientific findings by scientists and nonscientists alike.

This article considers replicability of the performance of predictors across studies. We suggest a general approach to investigating this issue, based on ensembles of prediction models trained on different studies. We quantify how the common practice of training on a single study accounts in part for the observed challenges in replicability of prediction performance. We also investigate whether ensembles of predictors trained on multiple studies can be combined, using unique criteria, to design robust ensemble learners trained upfront to incorporate replicability into different contexts and populations.