Hi everybody,

I am looking for direction on how to proceed with the following issue.

I am running some experiments relating to the appearance of traffic jams in a traffic simulation. The simulation is a multi-agent system with each car representing an agent. There are two ways that I measure the existence and extent of traffic jams in the environment.

The first looks at the number of streets with slow or stationary traffic. When throughput goes below a certain range the street is considered to have a traffic issue and as the traffic jam grows the number of streets in this state will increase. So by counting streets with this condition I get the first measure of traffic congestion in the system.

The second method is done by the agent cars. Without going into details, the cars indirectly measure congestion. Again, when the traffic around them goes below a certain threshold they say they in a traffic jam. As with the first method, the number of cars with this state increase or decreases as the traffic jam grows or shrinks.

So in the end I have two time series, one with the number of streets with congestion throughout the simulation and the second with the number of cars saying there is congestion. What I want to do is measure the effectiveness of the second method by showing that there is a relationship between the two series i.e. when the number of streets with congestion goes up, the number of cars saying there is congestion also goes up.

I've done some investigation and my idea is to use Granger Causality as a measure of this, but I am not convinced if this is valid. Any opinions on this or suggestions on other techniques to look into would be greatly appreciated.

Thanks in advance