This study investigated the impact of different degree of data aggregation on the travel time measurement study. For this study, the travel time measurement data were obtained from the GPS-equipped probe vehicle, Bluetooth devices, and Sensys detection system. The collected travel time data were controlled at three levels reflecting the degree of data aggregation: 1) averaged the results (two data points) from the GPS-equipped probe vehicle, 2) aggregated for entire period as a low level data reduction, 3) separated the period into ‘before the school starts’ and ‘after the school starts’ in order to consider the traffic demand derived from the school’s semester, and 4) only considered a specific time period selected based on identical traffic volume as a high level of data reduction. As a result, the GPS and Sensys-based before-and-after study had similar results indicating all westbound segments had travel time savings after the installation of SynchroGreen while a part of the eastbound segments did not. On the other hand, the Bluetooth-based study indicated the travel time savings on all eastbound segment, but not all on the westbound segments. In addition, even with the same data source, the Sensys-based results produced by separating the period were different in terms of the effectiveness of SynchroGreen when this comparison was based on the same volume condition and when it was not. Based on these results, this study emphasizes the importance of study design (i.e., data reduction) when a before-and-after study on the ITS technology is conducted.     «

This study investigated the impact of different degree of data aggregation on the travel time measurement study. For this study, the travel time measurement data were obtained from the GPS-equipped probe vehicle, Bluetooth devices, and Sensys detection system. The collected travel time data were controlled at three levels reflecting the degree of data aggregation: 1) averaged the results (two data points) from the GPS-equipped probe vehicle, 2) aggregated for entire period as a low level data re...     »