On Monday 1 March 2021, 4-6pm, the RSS Quality Improvement Section held a meeting, Dr Groundlove - or how I learned to stop worrying and love the Central Limit Theorem, where three speakers gave a thorough and entertaining review of a newly launched set of important environmental guidelines.
Peter Witherington is a director of RSK, an engineering and environmental consultancy based in the UK with over 5000 staff; Nigel Marriott is an independent statistician with extensive consultancy and training portfolio; and Ron Wasserstein is the executive director of the American Statistical Association, the ASA.
Peter wrote a report for CL:AIRE (Contaminated Land: Applications in Real Environments) in 2008 about ‘Guidance on Comparing Soil Contamination Data with a Critical Concentration’. He was then instrumental in setting up a review of this work that was undertaken by a steering committee with Nigel Marriott as the lead author. The update was published in 2020 and it was this new guidance that was the motivation for the event.
Peter explained how and why the industry needs to use statistics, the issues with the previous statistics guidance and why CL:AIRE decided new guidance was needed. He also demonstrated how the industry uses a Conceptual Site Model to identify which areas of the site need to be sampled further so as to make a decision on whether the contamination levels are severe enough for the site to be cleaned up.
Nigel’s mission is to explain complex statistical ideas to non-statistical audiences in such a way that they can make use of his results and it was this aspect of his career that he wanted to bring to the table when he was asked to lead the writing of the new statistics guidance for CL:AIRE. Nigel talked about the statistical content of the new guidance, and the work he did to satisfy himself that he had truly understood the limitations of the Central Limit Theorem and that his recommendations were statistically robust. Nigel’s biggest concern was that in trying to keep the guidance as simple as possible, he may have written a document that was too simple especially when the sample contains hotspots of high contamination. However, the guidance has been very well received and a successful balance has been achieved.
The ASA is the world’s largest community of statisticians and the oldest continuously operating professional science society in the United States. Since 2007, Ron Wasserstein has been responsible for ensuring that the ASA fulfils its mission to promote the practice and profession of statistics. He was the lead author of both the 2016 ASA statement on Statistical significance and p-values and the 2019 American Statistician editorial 'Moving to a world beyond P<0.05' and was therefore uniquely placed to critique the new guidance against these. Ron was delighted to have an opportunity to critique a piece of work that sought to implement the 2016 statement and 2019 editorial in a field that was not medical.
There was a wide discussion ranging from issues about data distributions to the state of statistical awareness in public policy sectors. One theme that did emerge was a general agreement that the contaminated land industry would be a good candidate for Bayesian statistical inference whereby the Conceptual Site Model introduced by Peter is used to define a prior. This could be particularly useful in situations where it is logistically and financially difficult to take a large sample.
Watch a video of the meeting.
Author
Report by Shirley Coleman, who chairs QIS, and is an applied statistician working at Newcastle University with special interest in knowledge transfer and engagement with business and industry.