My house gets super dry in the winter and I didn't understand why.
My weather station says it's drier inside than outside, which didn't make a ton of sense. I'm cooking, and sweating, and showering inside. It should be more humid.
I suspected my forced hot air system was to blame, but I didn't understand the mechanism. Cooling air makes it drier, but heating shouldn't remove any water. Like - I know my A/C has a drain on it, but my furnace definitely doesn't.
Relative humidity
As it turns out, I didn't understand my metric. My weather station reports relative humidity, and I guess I always just ignored the "relative" part?
Relative humidity is how close we are to the maximum humidity at a given temperature. Hot air can hold a lot of water, cold air holds very little water. The difference is huge - air at 68°F can hold ~4x more water than at 32°F (at least that's what I understand from skimming this wikipedia article).
So when I opened my windows to let in the 60% humidity 32° air, the relative humidity dropped to 15% after heating it to 68°. Mystery solved.
Beware shifting denominators
When a ratio moves, people tend to think it's because the numerator shifted. It can be very misleading when the denominator changes instead.
For example: the percent of users buying a product after visiting your store increases by from 25% to 50%! That's great! Unless you just lost a bunch of traffic to your store:
| Date | Purchases | Visits | Purchase Rate |
|---|---|---|---|
| Yesterday | 20 | 80 | 25% |
| Today | 15 | 30 | 50% |
Sure your purchase rate increased, but you made fewer sales overall. Damn the denominator!
"Be skeptical of ratios" is pretty good advice - at least make sure you understand how the denominator moves.
It's still useful
Some folks take this skepticism of ratios to an extreme and say you shouldn't use ratios when reporting data. That's a mistake.
You shouldn't make your reader work harder than they need to. A ratio is often the easiest way to describe what's happening in the data.
You have a responsibility to make sure the data aren't misleading, but hopefully your reader trusts you to guide them to the right conclusion. This is critical to good data storytelling.
Relative humidity wasn't useful for understanding why my house was dry. It is useful for measuring how dry the air feels. It would be maddening to report absolute humidity next to a saturation vapor pressure table instead.