Shinichi Nakagawa on Improving Statistical Reporting
Fast Moving Front Commentary, January 2012
Article: Effect size, confidence interval and statistical significance: a practical guide for biologists
Authors: Nakagawa, S;Cuthill, IC
Shinichi Nakagawa talks with ScienceWatch.com and answers a few questions about this month's Fast Moving Fronts paper in the field of Biology & Biochemistry.
Why do you think your paper is highly cited?
This is because the topic of our paper is how we can make scientific papers more informative and useful to readers, as a result of better interpretation of the statistical results. Therefore, every biologist should be interested in reading our paper and they could all benefit from our paper.
Does it describe a new discovery, methodology, or synthesis of knowledge?
Our paper describes a synthesis of knowledge regarding how we could improve our interpretation and presentation of statistical results.
Would you summarize the significance of your paper in layman's terms?
In many biological sciences, statistical reporting concentrates on whether results are "statistically significant" or not. However, such dichotomous decisions are misleading because statistical significance is always likely when a study has a large enough sample size.
When we interpret our scientific results, we should instead focus on effect size—for example, how much actual difference there was between the control and treatment groups. Differences are not present or absent, they are sufficiently large to be biologically important or sufficiently small and close to zero to be unimportant. This seems obvious, but the interpretation of effect size is often missing from published papers in biology.
Shinichi Nakagawa enlarging (i.e. paint with a special dye and drying an electronic dryer) a male house sparrow's bib size as a part of study to see the effect of male ornamentation in winter survival. The results are published in Nakagawa, S., Lee, J.-W., Woodward, B.K., Hatchwell, B. J. & Burke T. (2008), Differential selection according to the degree of cheating in a status signal, Biology Letters. 4: 667-669.
In our paper, we have provided practical guides for improving our statistical reporting (and software tools for the calculation of effect sizes) as well as reviewing the problems inherent in current practice.
How did you become involved in this research, and how would you describe the particular challenges, setbacks, and successes that you've encountered along the way?
I initially had a vague idea to write this type of paper when I was an undergraduate at the University of Waikato, New Zealand, in 2001. However, I didn't start working on it till I undertook my Ph.D. on parental care in house sparrows at University of Sheffield, UK, in 2003.
I met my co-author Prof. Innes Cuthill at University of Bristol, who happened to share my views on the topic, during my first field season (on Lundy Island, Bristol Channel, UK). It took us more than three years to write this paper as we had to work on this paper as a side project. The paper was eventually published in 2007.
Where do you see your research leading in the future?
I would like to further contribute to improving the use of statistics in biological sciences. Also, my recent interest is the use of computer simulations (especially agent-based modeling) in the field of ecology and evolution. I would like to popularize the use of agent-based modeling in my field.
Do you foresee any social or political implications for your research?
The correct interpretation of scientific results is indispensable when such results are used for social and/or political decisions. Our paper should improve the quality of scientific reporting in biological sciences in general. My co-author has been on a working group whose recommendations for scientific reporting, the ARRIVE guidelines, have been adopted by many of the world’s leading biomedical journals (view).
I am convinced that better science means better translation into practical solutions of benefit to society and more informed political decision-making.
Department of Zoology
University of Otago
Dunedin, New Zealand
KEYWORDS: BONFERRONI CORRECTION, CONFIDENCE INTERVAL, EFFECT SIZE, EFFECT STATISTIC, META-ANALYSIS, NULL HYPOTHESIS SIGNIFICANCE TESTING, P VALUE, POWER ANALYSIS, STATISTICAL SIGNIFICANCE, NONCENTRAL DISTRIBUTIONS, INFORMATION THEORY, ECOLOGY, SCIENCE, PSYCHOLOGY, EVOLUTION, SELECTION.