Trustworthy Statistics – Avoiding Deadly Sins, Statistical Lacunae and Public Misunderstanding
Sheila M. Bird
Biostatistics Unit, MRC, Cambridge, United Kingdom

Statistical science is study design and methods for analysing the resulting data. In medical statistics, progress was made a quarter century ago because clinical scientists became advocates for, and journal editors signed-up to, the importance of randomization, pre-specified primary outcome, power to discern plausible effect sizes, study protocols, structured abstracts, statistical guidelines for contributors to medical journals, and being wary about posthoc sub-group analyses and unplanned interim analyses.

How easily the unwary can mess-up statistically or the canny can beguile!

Yet, I am encouraged by 21st century efforts of scientists, civil servants, editors, journalists, press officers and parliamentarians to support the public's better understanding of statistics. I'll cite examples which range from swine-flu and criminal justice to military matters to illustrate the following themes.

Trust in statistics begins with robust experiments which both design-out bias and give precise answers, rather than second-rate studies which entrench bias or lack power.

Relied-upon statistics may not be worthy of the trust placed in them … with serious sequelae, just as serious as those medicine reacted to.

Absent statistics can tell a different tale.

Bizarrely conflicting statistics rightly rein-in decisions.

Statistical terminology is widely misunderstood …

Wherever in society data are, so too is recourse to statistical science.

The Royal Statistical Society (RSS) has recognised that journalists write well and has sought to celebrate and encourage their writing well about statistics. Straight Statistics, Sense about Science and RSS have collaborated on 'Sense about Statistics'. The Straight Statistics website highlights as 'statistical felony' knowing misuse of statistics; but also lauds answers to statistically-adept parliamentary questions which 'out' crucial data.

Brevity is demanded of researchers (by journal editors), of civil servants (by ministers) and in press-releases (to be newsworthy). I'll suggest seven deadly sins to be avoided – even in briefs!

Keywords: Statistical trust; Design; Public understanding

Biography: Professor Bird is senior scientist at the MRC Biostatistics Unit, and is a former member of RSS Council and a number of RSS working parties. She has served as vice president and honorary officer for external relations.