Summary of methodological issues in epidemiology
I’ve published a report prepared for Steve Baker MP into some of the issues I’ve seen in the past half year of reading papers and models from the world of epidemiology.
It can be found on my website here:
Summary of methodological issues in epidemiology
It discusses a mix of programming, statistical, logical and communication related issues with some suggestions to the government for how to better incorporate research into decision making.
Mr Baker has written an op-ed titled “I cannot support a second lockdown”. It’s unfortunately behind the Telegraph paywall, but he references the report in this paragraph:
Since the start of the outbreak, much of our policy has been model driven. I am on Wednesday releasing a report on methodological issues in epidemiology by Mike Hearn. On Tuesday, I asked select committee chairs to consider the issues. Models often contain internally inconsistent and non-replicable numbers, use insufficiently large data sets to derive critical inputs, and the bounds of uncertainty are often either not reported or have a very wide range. In one case, a prediction was changed to “20,000 deaths or much lower”. Academic software quality is notoriously poor, and models are often not validated by the actual course of an epidemic, yet they still drive policy. Models departing from the “accepted” viewpoint are often rejected for publication. Change is essential.
Steve then explains that he plans to take the unusual step of voting against the government and his own party today, against a second lockdown in the UK. Other MPs are expected to either abstain or join him.
This is occurring against a background of the British public and politicians losing trust in the scientific advice they are being given. Literally one day after Number 10 was shown predictions/projections/scenarios/whatever predicting overflowing hospitals within weeks, it was demonstrated that the idea was based on model outputs already weeks old and some of them had already been invalidated.
The first lockdown was meant to last three weeks. Parliament votes today on a second: it’s important that lawmakers have insight into some of the troubling things that have gone on within the scientific world in 2020.
I’ll keep an eye out for questions and may update this page later if some seem to crop up frequently.
