David Allison
๐ค SpeakerAppearances Over Time
Podcast Appearances
The example of ice cream consumption and murder rates is trotted out.
We say, you know, more ice cream consumption associated with more murder rates.
Guess what?
It's heat.
People eat more ice cream when it's hot.
They murder more when it's hot.
All true.
All fine.
But that's actually just a tiny piece of it.
There's so much more.
There's the measurement problem.
And that measurement is not random.
And even if it was random, it's usually not taken into account.
If it was random and we knew it was random and we took it into account statistically, we could make the problem kind of go away.
But it's often not random.
It may be correlated.
People who eat more of this may systematically bias their reporting down than people who eat less of that.
There's selection bias.
People may choose to be in the study or not choose to be in the study, and that may affect things.
There's what's called collider bias.