Subdivide one-hour bins

[Nate-Wessel]

We're leaving a lot of interesting data on the table by sticking with a one hour time bin whose limits are rigidly set a priori. A few thoughts:

• There is often much more data within an hour than is strictly needed to get a meaninful average
• Within an hour we would expect there to be substantial variation
• We can't currently observe that within-hour variation
• But we try to, badly, by looking at variations between one-hour averages, which as averages regress toward their mean.

I'd like to come as close as possible to observing the total travel times of single vehicles making a complete trip across the corridor. A single vehicle's complete travel time should be the fundamental unit of observation for any analysis of corridor-level travel times.

This would allow us to move away from only looking at variation between temporal averages to start to look more at variation per se and travel time reliability.