I used data the city of San Francisco to see if there was correlations between the number of traffic accidents and whether or not a bike lane was on a road. I overall conclude that there are proportionally more accidents on a bike lane than off a bikelane, in basically all districts. It's very possible this is because that the fact that bikers will stick to bike lanes instead of regular lanes is the reason for this.
- Bike lane Data (where bike lanes are, when installed, length of bike lane) was from DataSF
- Traffic Accidents data, was subsetted from a list of Police Reports, again from DataSF
01_DataCleaning.ipynb - This python workbook is the preliminary data cleaning work I have been doing in order to examine the impact of additional bike lanesin SF on accident rates. I use bike lane data, traffic reports filed by the police.
02_Analysis - preliminary charts to show how bike lanes vary with bike lane data, given time of day, neighborhood in SF, and year
`03_Analysis_StdMetric' - I realize in the previous analysis that there is a lot more roads without a bike lane than with a bike lane. I standardize the number of accidents by amount of road type (by bike lane length of group if on a bike lane vs regular road if not on a bike lane).
- I notice that while there are more accidents on a bike lane when I disaggregate by district, and/or by time of day
- A regression (with controls for year, length of road, length of bike lane ) estimates that there are more accidents off a bike lane than on a bike lane, and bike lanes decrease accidents. However the sample size for this regression was incredibly small