Hacker News new | ask | show | jobs
by alabaster_punt 2225 days ago
If you have a specific reason to believe that the model resulted in an incorrect conclusion, feel free to share it, instead of casting aspersions with cryptic nothings.
2 comments

The model said if we don't lock things down China-style millions will die. But we see now from Sweden it was not true. See top graph: https://informationisbeautiful.net/visualizations/covid-19-c...

Also Neil's other predictions have been completely wrong on the Swine Flu and Bird flu. Why this "expert in his field" is still listened to is a mystery to me.

If you're talking about that first graph on the page, I think that it is not very helpful. Comparing total number of deaths can be misleading, especially for small countries like Sweden. If my state had a Coronavirus death rate as high as Sweden's 15,000 more people would be dead. Still not "millions", but it wouldn't be an ideal outcome.
Weren't his projections for deaths assuming we did a lockdown also much higher than the actual numbers?
Some other comment here said it was the other way around,they predicted 20k deaths in the UK with a lockdown and the death toll is 36,000.
Sweden is pitched as as some kind of example of handling the pandemic correctly, when their numbers[1] don't look good at all. From their numbers[2] 44% of people diagnosed ends up dead, implying they are way under tested or they have ridiculously deadly strain (I think it is the former). Because of that the most reliable indicator from them is number of deaths.

Now compare number of deaths with countries that reacted quickly, for example: South Korea, Taiwan, Australia, New Zealand, Norway and others.

[1] https://covidgraph.com/

[2] https://www.worldometers.info/coronavirus/country/sweden/

It looked very wonky. I cannot judge if the code does not actually produce something worthwhile, but it certainly doesn't pass any smell tests. In no profession would work looking as sloppy as that be trusted. And for the critical decision it was used for I would expect something more robust.