As researchers we need to be more aware of the kind of research we are doing, and we need to be more honest and self-critical about these scientific works. Even more considering that later, based on the results of these studies, the so-called evidence-based practices are implemented, be they clinical, educational, health, political practices, among many others. Even more so considering that meta-analyses, positioned as the fundamental evidence when making decisions about a given field, base their findings on "primary" research (our individual publications).
In this article published in Nature ( https://www.nature.com/articles/d41586-019-01307-2 ), Dr. Dorothy Bishop of Oxford University highlights 3 factors that can do more harm to science: a) bias in publications (not publishing those data that do not indicate favorable results, or that are not statistically significant, for example); b) not considering statistical power in the design of studies (this leads some researchers to erroneously conclude that a hypothesis is rejected, see the excellent example recently used by Pascual Huerta: https://www.elsevier.es/es-revista-revista-espanola-podolog...); and c) the so-called "P-value hacking", which can be translated as those hypotheses that are generated after the researcher knows his results. Finally, the author mentions a phenomenon called "HARKing", a practice as widespread as the previous one and assumed as "a good, adequate, or normal way of proceeding" by many researchers. Although the author presents it as another factor, from my point of view "P-value hacking" and "HARKing" are intimately linked, given that the latter refers to the practice of looking at the results, making a finding that seems exciting, and writing an article around this result. That is to say, they are ways of proceeding post hoc, once the result is known, but nevertheless the researcher shows an article as if he had proceeded in reverse, that is to say, as if from the beginning he had a clear idea/hypothesis that he could later corroborate with his study.
It is a subject that is undoubtedly important to debate, and that touches on aspects that should not cease to be part of research teaching, along with method and statistics, more linked to ethics and moral.
Pablo - this is very true and yet study bias has been with us for as long as research has existed. Ben Goldacre's TEDs about clinical drug trials and EBP repositories (such as Cochrane Library) are a perfect illustration. For me, there are ways to limit this. One is multi-disciplinary team research. There is less likely to be a 'group-think' around manipulation of results and avoiding negative results. Mixed methods is another way around it (but again as team research) as multiple sets of data are collected - and it is difficult to manipulate data when triangulating.
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Publications