Research findings from one set of academics can be notoriously difficult to reproduce by others, leading to claims that up to 90% of research funding may be being wasted.
Research findings from one set of academics can be notoriously difficult to reproduce by others, leading to claims that up to 90% of research funding may be being wasted.
Now this so-called ‘reproducibility crisis’ has been put under the spotlight by researchers at the University, who have concluded that the quality of findings is primarily dependent on researchers’ critical approach to the data they include in their study.
They say that much greater understanding of the way academics behave in carrying out their research is required therefore if research funding is to be used more effectively.
Professor Martin Michaelis and Dr Mark Wass from Kent’s School of Biosciences, together with Professor Larry Ray from Kent’s School of Social Policy, Sociology and Social Research, analysed the currently published information on the ‘reproducibility crisis’.
They found that the discussion is largely focused on methodological issues, although the availability of established methods does not ensure good practice. In addition, research is often performed for the first time before standardised protocols are available. This suggests that the reliability of research findings primarily depends on the critical approach of researchers who generate the data and analyse them.
Professor Michaelis said: ‘Despite the crucial importance of the critical approach that researchers use towards their own data, evidence on this is largely lacking. Hence, meta-research is required to understand how researchers handle and present their data (including the selection of data that will or will not be included), how this influences data quality, and how this is affected by the research environment and its incentives.
‘The resulting knowledge will be essential if we are to improve data quality and the effective use of research funding’.
The study Understanding of researcher behaviour is required to improve data reliability and published in the journal GigaScience