If anyone is curious, my estimate of a potentially similar effect based on the weather would require a self-experiment of only ~70 days. I hope someone does run one.
EDIT: with additional data, the correlation has shrunk and so the necessary n has increased substantially to 1600 days.
Er… yeah. Confounders are why we would be interested in running a randomized experiment in the first place, and hence be interested in issues of experiment design like how many days of data we would need.
If anyone is curious, my estimate of a potentially similar effect based on the weather would require a self-experiment of only ~70 days. I hope someone does run one.
EDIT: with additional data, the correlation has shrunk and so the necessary n has increased substantially to 1600 days.
Weather has too many confounders.
Birdsong, for example, has quite apparent anxiolytic effects.
Er… yeah. Confounders are why we would be interested in running a randomized experiment in the first place, and hence be interested in issues of experiment design like how many days of data we would need.