We analyzed transportation survey data from the UC Davis community in which individuals were asked to rate their comfort level biking on certain streets based on 10-second videos of those streets.
We implemented Bayesian models with random effects to determine which features of streets and individuals had the strongest relationships with comfort ratings. Not surprisingly, we found a mix of street-level and individual characteristics to be important predictors. We found random effects to be important for controlling for individual tendencies to rank low or high, and for interactions between street-level variables that we couldn’t put explicitly in our models.
Our findings are being incorporated in a report for the National Center for Sustainable Transportation.