Abstract
Transparency is powerful in its simplicity. It evokes pleasant images such as a clear sky seen from the Alps or crystalline waters surrounding Mediterranean islands. To the politician, it is a reminder of their accountability; to the managing director, an obligation towards the company’s stakeholders; to the scientist, the principle without which science would be reduced to alchemy.
The Journal of Child Health Care (JCHC) has a readership that spans 192 countries in 6 continents (Antarctica, you are next). The journal features a broad scope of exciting research; all focusing on the health and healthcare of infants, children, young people and their families, but rooted in a range of different disciplines (from nursing and health services research to epidemiology, psychology and sociology). With the increasing geographical and disciplinary reach of modern day research, never has a shared, global language been more important. Scientific language offers the transparency and universality required for open communication, shared learning and the reproducibility of findings.
Many of the studies published in JCHC have relied on statistical analyses that, today, can be performed with freely or commercially available software on modest computers. For example, regression analysis has now become a familiar tool to any healthcare researcher who deals with multivariate data. However, just a few decades ago, simple multivariable models could only be estimated using a large and expensive, shared computer. Syntax had to be carefully written in advance using specialist programming languages, the job had to be logged in a queue and the analyst would wander off waiting several days for the model to run. In the past few years, we have seen the development of advanced regression techniques capable of coping with correlated data, missing values, non-normal errors, as well as non-linear associations. Not only are our statistical capabilities increasing but access to more advanced methods is improving.
The availability of low-cost computing power and user-friendly, specialized software is an important accomplishment that has greatly facilitated the spread of statistical methods. But, at the same time, the mathematics and assumptions behind these methods have been hidden away from the user and, as a consequence, the potential for statistical fallacies has increased. The time is ripe for a fresh look at modern statistics through the lens of basic principles in statistical reporting.
Statistical guidelines 1 are now available for authors who wish to submit their manuscripts to JCHC. These comply with the specific Journal’s style and, more generally, with conventions in statistical reporting. Instructions for presentation and terminology are designed to avoid common mistakes that might distract the reader from the narrative and negatively affect the outcome of the review process. Guidelines on methods provide both general and specific criteria against which the quality of the manuscript will be assessed. Moreover, the provision of detailed and exhaustive descriptions, in accordance with widely accepted guidelines as those provided by EQUATOR Network (http://www.equator-network.org/), is encouraged (either in the text or in the supplemental material to adhere to Journal’s word limit).
Research studies, from small clinical trials to large population-based surveys, require careful planning, understanding of sources of uncertainty and appropriate methods of data analysis. Integral to JCHC guidance is the recommendation that, when in doubt, statistical advice should be sought. This can help avoid pitfalls such as the deceitful Simpson’s paradox and the creeping ecological fallacy or the perils of ‘data dredging’ (Everitt and Skrondal, 2010). Timing, too, is important, as epitomized by an illustrious scholar: ‘To call in the statistician after the experiment is done may be no more than asking him to perform a postmortem examination: he may be able to say what the experiment died of’ (Sir Ronald Fisher, Indian Statistical Congress, Sankhya, ca. 1938).
There are clearly no absolute rules on how statistical methods and results should be presented. Nevertheless, any such rule needs to be based on transparency, which has inspired the formulation of these newly launched JCHC guidelines. After all, isn’t it nice to climb a mountain or dive into the sea and be able to enjoy the view?