Trajectory Distribution Metrics: Density

[JosephCottam]

Characterize a collection of trajectories to indicate how dense it is is in the span. This is a diagnostic of outputs, and may be used in a visualization (and other places).

This is related to statistical variance, but it needs to capture the variance of the variance over time. If all models give the same answer at some time, the density is 1. If the models span -inf/inf, then there is zero density. Reporting this density may be useful for a modeler to diagnose information.

To maximize utility, the density needs to be insensitive to the number of trajectories (though below a certain number, this is probably impractical).