3.3.3. Statistical Analysis

Because of the relatively small sample size, the three dummy variables, which depended on each criterion for classification, were used in both groups of the regression models. Altogether, four regression analyses were estimated. First, a dummy indicator of "before implementation" was used as a reference variable. In the change analysis, the VS-C or DC-C was regarded as a reference variable.

We controlled for the e ffects of factors expected to influence vehicle speed on shared streets (i.e., street width, tra ffic volume, and pedestrian volume), with the assumption that the narrower the street and the larger the volume of street users, the slower the vehicle speed. In addition, we controlled for tra ffic calming devices that might directly influence vehicle speed. The only tra ffic-calming device at the study sites was a speed bump, which we used as a control variable. To account for factors related to the filming location (intersection or mid-block), a variable, indicating the distance from the recording spot to the nearest intersection entrance, was included ("distance to closest intersection"). Last, dummy variables, indicating "morning" or "afternoon," were included ("evening" was the reference group) to control for the e ffects of the recording periods. The second regression analysis was about change and it used di fference (change) values instead of absolute values for the time-variant variables (i.e., " Δ tra ffic volume," " Δ pedestrian volume," and " Δ number of speed bumps").
