Cell Phone Data and Travel Behavior Research: Symposium Summary Report

The Exploratory Advanced Research Program

Relating Census Block and American Community Survey County Data, Demographic, and Economic Data to Derived Trip Data

In this session, Ben Pierce from Battelle presented on reducing bias and variance in sampling by linking cellular data with travel surveys and other datasets. Symposium participants then discussed challenges relating to changing methods based on emerging technologies.

Presentation

Ben Pierce, Battelle
Ben Pierce of Battelle began his presentation with an exercise to demonstrate the law of large numbers. This law indicates that as a group, it is possible to get close to the right answer, and as the size of the group increases, the closer to the answer the group will get. Pierce explained that statistics can measure variance in large surveys, as shown in figure 7; however, bias also needs to be considered. Pierce recommended taking an unbiased but imprecise survey and using precise cellular data to calibrate it. He noted that, even if cellular data are biased, it may still be necessary to accept a certain level of bias to reduce the mean squared error. 

Figure 7. Example of a large national survey with sample sizes in the tens of thousands.

Major Themes Discussed

The symposium participants discussed the following major themes after the presentation:

• Reconciling improvements. A key challenge moving forward is attempting to reconcile technological improvements with changing methods and how information is collected and analyzed.
• Reducing uncertainty. It is possible to reduce the level of uncertainty within the results.

Key Takeaways

During the discussion, the symposium participants agreed that leveraging and reconciling surveys and cellular data will be helpful in determining a broader picture of what is happening.