More Bad Data: CBC Ottawa's "Second Wave"

More Bad Data: CBC Ottawa's "Second Wave"

Last year I wrote a post about the Ontario Carbon Tax Stickers that dot the gas pumps across our province (link to that post: http://jordansworld.ca/the-problems-with-the-ontario-carbon-tax-stickers/) where I argued that it's the responsibility of governments to ensure the data they release is represented accurately. Since the SARS-CoV-2 pandemic began last year, there has been a slew of inaccurately modelled data released by governments, academia and the media.

I came across this graph, posted on CBC Ottawa's Twitter page on 19 June 2020:

Retrieved on 19 June 2020 from https://twitter.com/CBCOttawa/status/1273990355025879043/photo/2

The most glaring problem with this chart is the exponential yellow curve indicating what would happen if the current SARS-CoV-2 infection rate increased by 20%. This leads the viewer to believe that cases will begin rising exponentially starting in early July, and continue through October. Further, the viewer is drawn to the yellow curve since it is placed in front of the blue curve (the actual transmission rate).

Take note of the asterisk at the bottom right of the graph, '*shaded area represents the 90% credible region'. This is more or less a confidence interval, which is a term in used when calculating statistics that means 'we can expect the actual value to fall somewhere within this range 9 times out of 10'. In the CBC's chart, the confidence interval is literally 'off the chart' beginning in August, and drops down below the blue curve starting in late August. Essentially what this chart is saying is that there will be somewhere between zero and infinity (since they don't define an upper interval we need to assume it tends to infinity, as exponential functions do) COVID-19 cases in Ottawa hospitals by August. Given this very broad confidence interval, choosing a number between zero and infinity will produce a result that is as accurate (or possibly even more accurate, there are infinite possibilities after all) as the CBC's model.

Not only is the graph itself misleading due to the far reaching confidence interval, the source of the data is questionable as well. Following the link provided by the CBC (https://613covid.ca/) we arrive at a website that contains the model used by the University of Ottawa, The Ottawa Hospital and Ottawa Public Health. What is troubling is that in the model's parameters (https://613covid.ca/model/) we can see that the authors refer to Neil Ferguson's model, which has time and time again proven to be inaccurate. Ferguson's model predicted that, by June, 500,000 Britons would succumb to the virus. As I write this in late June, fewer than 45,000 Britons have succumbed to it, meaning Ferguson's model was off by a factor of more than ten. It is irresposible for the CBC to continue presenting data that follows this failed model since it is not painting an accurate picture of the current situation for the public.

Like I said in my previous post about the Carbon Tax Stickers, it is the duty of the media, governments and academia to strive to present the most accurate data possible to the public. Failing to do so erodes trust between them and the public, which hampers our ability to react to life-and-limb situations that may arise in the future. This is a cautionary tale of 'The Boy Who Cried Wolf', eventually the public will stop listening.

Show Comments