by Derek Winton
I should start by saying I don’t suspect any sinister intent on the part of Prof. Ferguson or believe he’s part of a conspiracy. As someone with a background in the hard sciences who also got their start in the heyday of British ‘bedroom coders’ I even feel a certain amount of kinship.
To me this is a governance issue. Decisions on pandemic response strategies affect millions of lives and in my (hopefully not unreasonable) opinion should be based on the very highest quality of information and subject to the highest level of scrutiny.
In my article I made eight substantive claims, summarised below:
1. The Imperial model was influential in the decision to pursue a lockdown strategy.
2. The research for ‘Report 9’ was not peer reviewed.
3. The model was not documented.
4. Prof. Ferguson apparently has no formal training in computer modelling, medicine or epidemiology.
5. Projections of death tolls from the same team in previous epidemics had been out by several orders of magnitude.
6. The code was of poor quality from a legibility stand point.
7. The model is an attempt to model a highly complex (and therefore highly sensitive) system but omits at least one key variable.
8. Projections based on the Imperial model for Sweden were out by a factor of 7 and therefore the model was not fit for purpose.
Leaving aside the link to a Conspiracy Theory Handbook, Prof. Ferguson does not appear to dispute any of these points. Instead he points out that other models made equivalent predictions and the report I cite for predictions on Sweden did not in fact use the Imperial model. In the absence of further information, we can only conclude that they are all in fact true. It’s tempting to say at this point that the prosecution rests, but of course we should allow Prof. Ferguson the chance to rebut any of the above points.
Prof. Ferguson does raise some points though that raise even more questions. Taking them in turn:
The Imperial model was influential in the decision to pursue a lockdown strategy
Prof. Ferguson points out that several other models upon which the Government relied “all agreed”:
The models written by LSHTM, Warwick University and Institut Pasteur Paris all agreed with “the” Imperial model. All used different code bases. And in fact, there was never “one” Imperial model, but several. We now have 4 different COVID models, again which all agree.
Where are the code bases, designs, documentation and assumptions for these models? Given that the Imperial model was considered the gold standard, couldn’t we be forgiven for having concerns about these models too?
What exactly is meant by the term ‘agree’? Do they predict the same death tolls in all of the scenarios modelled in report 9? If not, by how much do they vary?
Prof. Ferguson also states that :
Government responses … were driven by the reality that any disease which generates epidemics which double every 3-4 days and for which over 2% of those infected require hospitalisation will overwhelm any health system that exists[Emphasis mine]
This is what philosophers would call question-begging, i.e. assuming what we are trying to prove. Of course if we assume the epidemic doubles every 3-4 days indefinitely, any health service would be overwhelmed. The critical question for the government was to determine whether the epidemic would continue to double every 3-4 days and for how long, and it was this question that the computer modellers purported to answer.
Projections based on the Imperial model for Sweden were out by a factor of seven
Prof. Ferguson claims that “no-one ran the Imperial model for Sweden (other than us)”. Here he is absolutely correct. Indeed it is impossible for anyone outside of the Imperial team to run the exact model (used to generate Report 9) as the original source code was never released.
It’s surely reasonable to ask then: If the team did indeed model Sweden, what did they find?
Derek Winton is the prospective parliamentary candidate for Lothian Region for the Reform UK Scotland in the forthcoming Scottish Parliament election on May 6th. You can follow him on Twitter at @derekwinton.