Carole Goble on reproducible research: what it really means & how to reach it

Posted by Biome on 8th July 2014 - 5 Comments


At the heart of scientific method lies the often elusive concept of reproducibility. It is increasingly apparent that much data churned out of labs across disciplines doesn’t meet this crucial criterion. Getting to grips with how to tackle this problem requires a firm understanding of exactly what is required to attain the badge of reproducibility. What is really meant by the catchphrase ‘reproducible research’? What is needed to practically achieve this? In this video Q&A Carole Goble, Editorial Board member for GigaScience, offers some much needed answers.

 

“There is some confusion about what people mean by reproducibility […] Do they mean its repeatable? Do they mean its replicable? Do they mean its reproducible? Do they mean its reusable? These are quite distinct concepts.”
Carole Goble, University of Manchester

 

Goble is Professor of Computer Science at the University of Manchester, UK, where she carries out research into the design, development and use of data and knowledge management systems. She also leads the UK’s largest e-Science pilot project, myGrid, is a co-investigator at the Software Sustainability Institute, UK, and co-director of the UK e-Science North West regional centre. Her research interests include the Semantic Web, medical informatics, e-Science, Grid computing, and bioinformatics. Here Goble shares her thoughts on what’s needed to make reproducible research a reality.

 

“You need at least four things to be able to reproduce something. […] The fourth thing that nobody does properly, in my experience, is have a clear description of the degree of variation and boundaries of tolerance.”
Carole Goble, University of Manchester

 

For more from Carole Goble, read what she has to say about the importance of truly open data.

  • Nolan Nichols

    I see quite a wide variety of definitions for research that is “repeatable”, “replicable”, “reproducible”, and “reusable”… Are Carole’s definitions written down somewhere that is easy to reference?

  • Pingback: July 21, 2014 | kuspfyi

  • Ged Ridgway

    I like the distinction between repeatable (same lab), replicable (different lab but same/similar context), reproducible (same/similar result from varied data/context/methods), and reusable (tools and/or data available). However, I think it’s awkward to then say “reproducibility means all these things”, since a result can clearly be reproducible (with some work) without its data or tools being directly reusable. Perhaps better to use the term “generalizable” for the specific use of reproducible, thus leaving “reproducible research” to refer vaguely to the whole set?