Read on Twitter
I've heard this come up frequently ("restaurants only 1% of identified cases"). It was said by Dr. Hinshaw and repeated by Premier Kenney as argument against further restaurant restrictions.
A thread on why I think this data point is misleading... https://twitter.com/tylerrdawson/status/1326921718045179904
A thread on why I think this data point is misleading... https://twitter.com/tylerrdawson/status/1326921718045179904
We often (always?) have to make inferences from a sample of the full population of data. In this case we know (IIRC) roughly 1/3 of cases are from private social gatherings, 15% from workplaces, 1% from restaurants... BUT roughly half are of unknown origin.
The key to making inference from a sample is whether the answer to the question: "is my sample representative of the whole population?" true or not.
In this case, do we believe unknown origin cases are split similarly as the identified ones?
I believe not. Here's why.
In this case, do we believe unknown origin cases are split similarly as the identified ones?
I believe not. Here's why.
Take 2 people: Paul (went to private party) & Judy (restaurant). Both test +ve. Contacted by AHS Tracer, Paul's able to list names and phone #s of all attendees. AHS is able to confirm cases from the party. AHS calls restaurant Judy went to, which who has no record of the diners.
It seems plausible that private gatherings and workplaces, where tracing is easier, will be more easily traced and identified. It's also plausible that a big chunk of the ~50% unknown origin cases are from hospitality and other public places, where tracing is more challenging.
What does data outside AB tell us?
A recent US study using 98 million cell phone records shows that restaurants, gyms, and other indoor venues were a disproportionate share of infections in the early months of the pandemic. https://www.nytimes.com/2020/11/10/health/covid-indoor-venues-infections.html
A recent US study using 98 million cell phone records shows that restaurants, gyms, and other indoor venues were a disproportionate share of infections in the early months of the pandemic. https://www.nytimes.com/2020/11/10/health/covid-indoor-venues-infections.html
We are all aware that many small businesses are teetering financially. No one wants to tilt them over the edge.
But relying solely on identified data, without considering potential statistical bias in unknown cases, to make public health decisions is potentially problematic.
But relying solely on identified data, without considering potential statistical bias in unknown cases, to make public health decisions is potentially problematic.
So if you hear Premier Kenney at today's press conference repeat the "restaurant risk is low" statement, please do bear in mind that's a conditional statement: conditional on only using data from what's been identified. Logic and US data suggests it may not be true. /end
