To build a reliable and accurate microsimulation model, it is important to calibrate to a single observed-representative day, rather than the average of multiple days of data. An average of multiple days could create unrealistic performance measures. Additionally, averaging creates the impression of weak bottlenecks with inconsistent time-dependent flow rates. To select a representative day from the travel conditions, key performance measures used in the cluster analysis are averaged at 15-minute time intervals across all days the data was collected. Then the percentage difference between the average and the observed is calculated. Based on this calculation, the representative day is selected, which is the day with the lowest percentage difference. The process of selecting a representative day is typically completed by using a cluster analysis, which helps account for variations in data caused by weather, incidents, and construction. For additional guidance on how to complete a cluster analysis, refer to Chapter 2 of the FHWA Traffic Analysis Toolbox, Volume III.

An

example

of selecting the representative day from a single cluster is provided in

Appendix N, Section 3 – Calibration Criteria Calculations Example

.