Tuesday, 1 September 2020

Lies and Damn Lies Applied to Statistics?

 


Statistician David Spiegelhalter has reacted bullishly to a writer in The Daily Telegraph who has warned that there will be an "inquiry into the role of statisticians in the Covid-19 crisis (https://www.theguardian.com/commentisfree/2020/aug/31/politicians-covid-19-statistics-statisticians). He cites Nate Silver who wrote that "numbers have no way of speaking for themselves.....we imbue them with meaning." Spiegelhalter is dismissive of the UK government's 'number theatre' in the early days of the pandemic, when there were daily presentations of numbers accompanied by 'fancy' graphs. Subsequent events have revealed that the numbers can be changed and that politicians generally  'pick and mix' the data, to support the line that appeals to them. There have been some almost laughable misuses of numbers, including a recent statement by Stephen Hahn (Commissioner of the US Food and Drug Administration). He claimed, at a press conference, that if treatment with the antibody-containing plasma from recovered patients was given to "100 people who were sick with Covid-19, 35 would have been saved". This remarkable 35% 'cure' is based on a finding that 8.9% of the plasma-treated group died compared to a 13.7% mortality in the 'control group' (who didn't receive the plasma). That is a 35% relative difference but the absolute difference is only 5%. This was also based on a single somewhat dodgy (scientists call such things 'preliminary') study, which wasn't even a randomised design (with the groups carefully matched to include people of similar age, gender, ethnicity, underlaying health conditions and disease severity). A 5% saving is distinctly optimistic. The question that remains is, do politicians 'twist' the data simply because they are ignorant or do they calculate what they can get away with to tell 'their' story? Either way, I don't think they are serving their populations well. Viruses don't care!

No comments:

Food For Thought?

The link between global heating and food prices is clearly illustrated in a recent CarbonBrief ( https://www.carbonbrief.org/five-charts-ho...