More readings, a better mean

Random errors make readings scatter, but their mean settles down as the readings grow in number. A systematic error does not. You will make the program take $3$, $10$, $100$ and $1000$ timings of a swing whose true period is $1.500\ \text{s}$, and watch what happens.

The idea. The mean is the best value, and the mean absolute error says how far a single reading usually is from it:

Your tasks.

1. Make mean(values) return the mean.
2. Make mae(values) return the mean absolute error.
3. Every timing is $0.04\ \text{s}$ too long because the watch is stopped late. Find the mean of these late timings.