Sponsored by

COVID-19 – Why I Model

OPINION

Like everyone, I’m scared.

I’m scared of the unknown, of the uncertainty and of the lack of leadership that is being shown in this crisis that is unfolding throughout the world.

As an engineer with a knowledge of statistics and economics, I spend much of my time staring at tables and graphs. As a scientist, my role is to interpret and synthesis data quickly and efficiently to reach valid conclusions.

In this time of great uncertainty, I decided to put my skills to use and start tracking how the COVID-19 virus has been developing throughout my state of South Australia. Every day I have been watching the press conferences live from my state government’s health agency SA Health and I have been analysing the data they have been making publically available.

My Model

As an engineer, I know that we can use trade-offs to analyse situations. I’m not an expert in virology. I did, however, learn the basics of how viruses travel through populations using a “good enough” model in highschool mathematics. Thankyou Mrs Kennedy.

I’m not saying my model is perfect. I’m not saying that this basic model is what the health care sector should use to make predictions.

Don’t be absurd.

What I am saying is that the simple exponential models are “goldilocks“. Just like the porridge in the story of the three bears… the model is good enough.

My — good enough — model is a simple exponential equation plotted on a log scale with three trend lines to show the rate of growth should it be doubling every 3, 4 or 5 days. As the infected sample comes in on a daily basis from SA Health the number is plotted on the figure and we can track, in a rudimentary fashion, how the state is growing while under lockdown measures.

Of course, this tool is only as good as the data that goes into it. This tool is highly reliant on accurate testing. Something which has been increasingly under scrutinisation in recent days. There are limitations. If you’re looking for something that has no limitations, you don’t understand science.

Importantly, because I am using an engineering approach to my data analysis and not a medical one, any conclusions should be taken to only be (at best) 80% accurate. Standard requirements for medical should be closer to 99.95%. I am not bothered with a rigorous error analysis (if someone wants to please do!).

I am sacrificing accuracy for speed to enable quick decision making in this time of rapidly evolving data.

Ultimately, this model is designed to provide some form of error checking against the political spin being produced in my State. That is all.

As Don Chipp said, “keep the bastards honest”.

Reception

Many of the pubic, my friends and family, have liked what I am doing.

Many others have criticised what I am doing as unethical – one can only assume for acting in a field that I am not an expert in.

Ultimately, I am engaging a course of scientific enquiry based on skills I have been developing since high school year 10 maths on exponential equations. That is more years than I would like to admit.

My analysis is not for everyone. It is only one perspective among millions. I understand that some may see what I am doing as driving “fake news”. Critically, what has happened is that my analysis is providing another perspective on data which is otherwise not being explained by our Government in South Australia.

Impact

We have some critical questions that need answering in South Australia.

Is the testing regime that is being touted as world best merely a triage test or is it a true reflective sample of the population?

Why is the testing criteria still limited in such a way NOT to find community transmission?

By having this data, we can attempt to answer these questions; or provide a starting point for those with more expertise to do so.

For example, by taking data provided by our state’s opposition regarding the number of tests conducted and correlating that to the number of positive samples returned we can get a better proxy for how the virus is trending in South Australia.

Fundamentally, the growth based on this proxy remains stable… it is not decreasing as is currently being reported in the media. This would indicate that the public perception that we are not testing enough (or testing the right people) would be accurate from a sampling perspective.

Now as a Veteran, I understand the need to make decisions in a time of crisis based on various inputs. I empathise with our leaders who are trying to monitor the growth in our State while conserving our limited supplies. There is a global shortage of the reagent needed to conduct the test. But, the situation needs to be presented correctly.

Since publishing my non-expert opinion (I’m not a virologist and I make no misrepresentations thereon) our State Government here in SA has started releasing graphs and data in their own dashboard. This is a step in the right direction.

https://www.sa.gov.au/covid-19/dashboard

Since publishing my own analysis, SA Health has also now reported numbers removing the effect of cruise ships in out State to focus on the domestic case. This was something I always have done.

Is my model being successful in “keep(ing) the bastards honest?” I like to think so. I’ve noted a significant uptick in media and political advisors from my state following me in recent days. That might, of course, be because I’m becoming a viral sensation as everyone laughs at how inaccurate my data is!

Is my model providing a sense of relief for me in this time of pandemic? yes. There is clarity in numbers.

To my detractors (of which there has only been a few), if you notice something incorrect come together and work on it together. Many of you have and I have learned much from your input. To the few who have simply said I should stop I would ask you to look into your own feelings and ask why? Is it your own fear which is driving such a reaction? In a time of crisis such as now, scientific and academic enquiry should continue. Fear and ignorance is never a reason to silence academic freedom.

Attempting to use ethics as a weapon to do so is abhorrent and should always be called out.


I update my model daily and publish to Twitter. You can view the current analysis at @rhematt.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.