Originally Posted by
KenHamer
Delays cannot follow a normal distribution, as there is a lower bound. That is, a flight cannot be delayed less than 0 seconds; a flight delayed by a negative time is not a delayed flight - it arrived early. Meanwhile, there is no upper bound to how long a flight can be delayed (particularly with Air Canada.)
You are looking for something more like a Poisson distribution.
Technically correct, but I didn't want to get overly complicated. Use a log-normal distribution if you crave for precision. But for simplicity, take a population of arrival times (ignore the actual schedule), calculate the mean and standard deviation, add two standard deviations to the mean and you'll get the latest probable arrival time with 97.5% confidence. Or close enough for all practical purposes, as we say in the engineering world.