- Let’s get this out of the way: Reproducibility is a cornerstone of science: Bacon, Boyle, Popper, Rheinberger
- A ‘lexicon’ of reproducibility: Goodman et al.
- What do we mean by ‘reproducible’? Open Science collaboration, Psychology replication
- Reproducible – non reproducible – A false dichotomy: Sizeless science, almost as bad as ‘significant vs non-significant’
- The emptiness of failed replication? How informative is non-replication?
- Hidden moderators – Contextual sensitivity – Tacit knowledge
- “Standardization fallacy”: Low external validity, poor reproducibility
- The stigma of nonreplication (‘incompetence’)- The stigma of the replicator (‘boring science’).
- How likely is strict replication?
- Non-reproducibility must occur at the scientific frontier: Low base rate (prior probability), low hanging fruit already picked: Many false positives – non-reproducibility
- Confirmation – weeding out the false positives of exploration
- Reward the replicators and the replicated – fund replications. Do not stigmatize non-replication, or the replicators.
- Resolving the tension: The Siamese Twins of discovery & replication
- Conclusion: No scientific progress without nonreproducibility: Essential non-reproducibility vs . detrimental non-reproducibility
- Further reading
There is a lot of thinking going on today about how research can be made more efficient, more robust, and more reproducible. At the top of the list are measures for improving internal validity (for example randomizing and blinding, prespecified inclusion and exclusion criteria etc.), measures for increasing sample sizes and thus statistical power, putting an end to the fetishization of the p-value, and open access to original data (open science). Funders and journals are raising the bar for applicants and authors by demanding measures to safeguard the validity of the research submitted to them.
Students and young researchers have taken note, too. I teach, among other things, statistics, good scientific practice and experimental design and am impressed every time by the enthusiasm of the students and young post docs, and how they leap into the adventure of their scientific projects with the unbent will to “do it right”. They soak up suggestions for improving reproducibility and robustness of their research projects like a dry sponge soaks up water. Often however the discussion is in the end not satisfying, especially when we discuss students’ own experiments and approaches to research work. I often hear: “That’s all very good and fine, but it won’t get by with my group leader.” Group leaders would tell them: “That is the way we have always done that, and it got us published in Nature and Science”, “If we do it the way you suggest, it won’t get through the review process”, or “We then could only get it published in PLOS One (or Peer J, F1000 Research etc.) and then the paper will contaminate your CV”, etc.
I often wish that not only the students would be sitting in the seminar room, but also their supervisors with them! Continue reading
I failed to reproduce the results of my experiments! Some of us are haunted by this horror vision. The scientific academies, the journals and in the meantime the sponsors themselves are all calling for reproducibility, replicability and robustness of research. A movement for “reproducible science” has developed. Sponsorship programs for the replication of research papers are now in the works.In some branches of science, especially in psychology, but also in fields like cancer research, results are now being systematically replicated… or not, thus we are now in the throws of a “reproducibility crisis”.
Now Daniel Fanelli, a scientist who up to now could be expected to side with those who support the reproducible science movement, has raised a warning voice. In the prestigious Proceedings of the National Academy of Sciences he asked rhetorically: “Is science really facing a reproducibility crisis, and if so, do we need it?” So todayon the eve, perhaps, of a budding oppositional movement, I want to have a look at some of the objections to the “reproducible science” mantra. Is reproducibility of results really the fundament of scientific methods? Continue reading
It is for good reason that researchers are the object of envy. When not stuck with bothersome tasks such as grant applications, reviews, or preparing lectures, they actually get paid for pursuing their wildest ideas! To boldly go where no human has gone before! We poke about through scientific literature, carry out pilot experiments that surprisingly almost always succeed. Then we do a series of carefully planned and costly experiments. Sometimes they turn out well, often not, but they do lead us into the unknown. This is how ideas become hypotheses; one hypothesis leads to those that follow, and voila, low and behold, we confirm them! In the end, sometimes only after several years and considerable wear and tear on personnel and material, we manage then to weave a “story” out of them (see also). Through a complex chain of results the story closes with a “happy end”, perhaps in the form of a new biological mechanism, but at least as a little piece to fit the puzzle, and it is always presented to the world by means of a publication. Sometimes even in one of the top journals. Continue reading
Based on research, mainly in rodents, tremendous progress has been made in our basic understanding of the pathophysiology of stroke. After many failures, however, few scientists today deny that bench-to-bedside translation in stroke has a disappointing track record. I here summarize many measures to improve the predictiveness of preclinical stroke research, some of which are currently in various stages of implementation: We must reduce preventable (detrimental) attrition. Key measures for this revolve around improving preclinical study design. Internal validity must be improved by reducing bias; external validity will improve by including aged, comorbid rodents of both sexes in our modeling. False-positives and inflated effect sizes can be reduced by increasing statistical power, which necessitates increasing group sizes. Compliance to reporting guidelines and checklists needs to be enforced by journals and funders. Customizing study designs to exploratory and confirmatory studies will leverage the complementary strengths of both modes of investigation. All studies should publish their full data sets. On the other hand, we should embrace inevitable NULL results. This entails planning experiments in such a way that they produce high-quality evidence when NULL results are obtained and making these available to the community. A collaborative effort is needed to implement some of these recommendations. Just as in clinical medicine, multicenter approaches help to obtain sufficient group sizes and robust results. Translational stroke research is not broken, but its engine needs an overhauling to render more predictive results.
Read the full article at the Publishers site (STROKE/AHA). If your library does not have a subscription, here is the Authors Manuscript (Stroke/AHA did not allow me to even pay for open access, as it is ‘a special article…’).
Recently, NIH Scientists B. Ian Hutchins and colleagues have (pre)published “The Relative Citation Ratio (RCR). A new metric that uses citation rates to measure influence at the article level”. [Note added 9.9.2016: A peer reviewed version of the article has now appeared in PLOS Biol]. Just as Stefano Bertuzzi, the Executive Director of the American Society for Cell Biology, I am enthusiastic about the RCR. The RCR appears to be a viable alternative to the widely (ab)used Journal Impact Factor (JIF).
The RCR has been recently discussed in several blogs and editorials (e.g. NIH metric that assesses article impact stirs debate; NIH’s new citation metric: A step forward in quantifying scientific impact? ). At a recent workshop organized by the National Library of Medicine (NLM) I learned that the NIH is planning to widely use the RCR in its own grant assessments as an antidote to JIF, raw article citations, h-factors, and other highly problematic or outright flawed metrics. Continue reading
Using metaanalysis and computer simulation we studied the effects of attrition in experimental research on cancer and stroke. The results were published this week in the new meta-research section of PLOS Biology. Not surprisingly, given the small sample sizes of preclinical experimentation, loss of animals in experiments can dramatically alter results. However, effects of attrition on distortion of results were unknown. We used a simulation study to analyze the effects of random and biased attrition. As expected, random loss of samples decreased statistical power, but biased removal, including that of outliers, dramatically increased probability of false positive results. Next, we performed a meta-analysis of animal reporting and attrition in stroke and cancer. Most papers did not adequately report attrition, and extrapolating from the results of the simulation data, we suggest that their effect sizes were likely overestimated. Continue reading