Hacker News Reader

The comments on the post suggest there are some major flaws. One graph compares two different rates on the same axis, and some others show small difference in logarithmic scales.

Can anyone look into these criticisms and see if they hold weight?

View on HN

What that comment is saying is that the two axis have different units (a unitless risk for covid case fatality and a time^-1 rate for mortality), so you can't compare them. Which is true, but if you multiply the rate by 1 year, then the units become the same and the values don't change, so that becomes comparable. I still think it makes sense to compare the risk of dying of covid-19 if you catch it vs the risk of dying of any cause in the next 12 months. 1 to 2 years worth of risk still seems like a lot to me

Right, and plotting the comparison to 1 year's risk seems entirely sensible to me. That seems approximately the right "compared to what?" scale to be using, otherwise almost any numbers multiplied by 300 million end up looking huge.

One thing this plot does very well is put into perspective the extra risk to the elderly.

Valid criticisms. Linear plot has been added but doesn't resolve the unit difference. The whole article tries to ask the right question but goes down the wrong path.

Can you elaborate in what way it goes down the wrong path?

Reply on news.ycombinator.com