Models are not data!

Catastrophically erroneous epidemiological modelling by Imperial College, London was responsible, early in 2020, for creating much of the panic that has since characterised Covid policy, here as well as in the UK. Furthermore, IC’s Prof Ferguson had a woeful record of similarly unsuccessful ‘arithromancy’ (BSE, SARS, FMD), none of which appears to have diminished the UK govt’s confidence in his work,

All this is worth remembering as, here in Australia, we seem to be witnessing an outbreak of low-grade modelling warfare, with rival institutes contending for the government’s ear. So far, it appears that the Doherty Institute enjoys government favour, that its forecasts are some of the most optimistic on offer, and that ScMo has had the good sense and moral fortitude to reject the wilder prognostications, such as that by the group headed by Prof Zoe Hyde of UWA, which warned that:

“…if Australia reopens once 80% of adults are vaccinated, which translates to 65% of the population overall, there could be approximately 25,000 fatalities and 270,000 cases of long Covid.

If Australia reopens with 80% of adults vaccinated and all children vaccinated, estimated deaths would fall to 19,000, or to 10,000 if 90% of adults were vaccinated.

In contrast, the Doherty Institute tweeted

In the COVID-19 modelling, opening up at 70% vaccine coverage of the adult population with partial public health measures, we predict 385,983 symptomatic cases and 1,457 deaths over six months.

8:31 PM · Aug 23, 2021

While this is some comfort to lockdown sceptics, these are, nonetheless, numbers that will elicit considerable push-back from the more persistently risk-averse among us.

I thought it reasonable to ask the Doherty Institute for evidence, in the form of case studies, that demonstrates the predictive skill of their methods, now that they are informing our govt’s Covid response.

Sent: Tuesday, 24 August 2021 11:45 PM

To: Doherty Reception <doherty-reception@unimelb.edu.au>

Subject: [EXT] Covid modelling

External email: Please exercise caution

Dear Sirs,

Your epidemiological modelling is currently being used to inform Covid policy in Australia. Can you refer me to earlier work of this kind that you have conducted using similar methodology, that demonstrates the predictive skill of the algorithm and methods you are using?

I look forward to hearing from you,

Kind Regards,

Tom Forrester-Paton

After a couple of days, I got the following boilerplate reply:

Dear Tom,

Thank you for your email.

Due to the volume of enquiries the Doherty Institute is currently receiving, we are unable to reply to individual emails at this time.

If your matter relates to COVID-19, please visit the Department of Health website in the first instance.

Thank you,

Kind regards,

Katie

Katie (she/her)

Receptionist

T +61 (0) 3 903 53555

katie.neumann@unimelb.edu.au

The Peter Doherty Institute for Infection and Immunity

792 Elizabeth Street | Melbourne | Victoria | Australia | 3000 doherty.edu.au

The Doherty Institute acknowledges the Traditional Owners of Country throughout Australia. We pay our respects to Elders, past, present and future.

Touched by their respect for the Traditional Owners of Country, but still anxious to be reassured that Australia’s Covid policy was not being driven by numbers plucked from thin air, I have just sent the following response:

Dear Katie,

Thank you for your reply. I appreciate that your Covid workload may be high, but all I am asking for is a link to a case study which, given the importance of this work to public policy, should be readily to hand, and the work of seconds to forward to enquirers.

I have looked at the Department of Health website, and can find nothing that addresses my concerns.

Given the momentous impact of the policies which your organisation’s modelling is informing, the need for transparency in demonstrating its skill should be obvious. With this in mind, I urge you to respond to my request.

Kind Regards,

Tom Forrester-Paton

I’m not holding my breath…

2 thoughts on “Models are not data!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s