Statistics don't lie. But people lie using them.
Below is what I call lying with the truth. It's done to sell newspapers of course; I mean that's a very alarming increase in Omicron deaths isn't it? Wow, just look at that steep bold line.
But hey, wait a minute. Let's take a closer look at the scale here. This is "Cumulative Deaths per 100,000" across various countries, with the USA looking much much worse than the other selected countries. From about the beginning to the end of January, deaths tripled. Three times anything is a lot, isn't it? Well, no. Not necessarily.
But hey, wait a minute. Let's take a closer look at the scale here. This is "Cumulative Deaths per 100,000" across various countries, with the USA looking much much worse than the other selected countries. From about the beginning to the end of January, deaths tripled. Three times anything is a lot, isn't it? Well, no. Not necessarily.
What sins did the Times commit with this chart?:
- Spuriously matching the boldface & misleading headline to the alarming "U.S." line, both in bold.
- Using the misleading term "per capita" in the headline while the more accurate "per 100,000" is much smaller and greyed out.
- Designing the scale so 30 deaths appears much more alarming than 10 deaths.
- Failing to tell us whether these are Omicron deaths or total COVID deaths.
But the last bullet point doesn't even matter. Here's my point: As serious as any 30 deaths are (especially to those 30 people and their friends and families), out of 100,000 people they are no more statistically significant than 10 deaths. Out of 100,000 people, 30 is still so close to zero as to be as statistically meaningless as 10.
This shows the danger of the Summary Metrics error or how one chooses to display data.
