Hacker News Reader

Comment by brookst | Hacker News Reader

Yep. That 30% is a bad use of statistics.

28 degrees Celsius is not 30% warmer than 21 degrees Celsius. This same stat rendered in Fahrenheit would say 70 degrees -> 82 degrees, or 17%. In kelvin it would be 294 -> 301, or 2.3%

Or we could invent a new measure indexed to Celsius but offset by 20 degrees, and declare a 1 -> 8 change, a whopping 700%.

View on HN

You see the opposite effect with reporting on the DJIA, where a 500 point move is treated like a big event even though it's is not nearly as significant today as it was 30 years ago. They ignore the more relevant percentage change in favor of the more sensational representation.

That's not statistics as much as sensationalism. Every morning on WSJ at about 7am you'll see a 'live' update about how the markets and futures are in a 'selloff' or 'surging' even if the percentages are like 0.3% in either direction.

Even Kelvin is the wrong model. What we need is temperature, humidity and air speed to interfere anything meaningful. ISO 7730 is even more precise. Any meaningful discussion should use the models from there.

But 28C or 82F is swelteringly hot in Britain so it kinda makes sense even it's incorrect.

Yeah. It's obviously incorrect in the sense celsius has no meaning as an absolute temperature scale and so its not 30% more anything, but in terms of colloquial meaning the average Brit probably does see it as ~30% further along an indoor temperature scale from "someone put the heating on please" to "crikey, it's sweltering in here"

This is not okay. If it makes sense for you, your sense for percentages is wrong.

Reply on news.ycombinator.com