Open terryf82 opened 6 years ago
think a possibility here is using Public Works data. This is available for Boston and likely other cities.
I think the plan is to handle this task now with our 'simulated intervention' idea (adjust a specific property/properties on a segment and display the impact) rather than go after actual interventions or looking at other segments. @bpben have I got that right?
Answering @terryf82 's question: This is where we got to with the intervention effect exploration: 8292c73d61788d3b5008b093d29186632f4a4b5a. We're going to pause on this for now because it did not seem particularly compelling after talking to BTD.
This is a core project goal and may not be solely a 2.0 feature, but BTD have indicated that being able to estimate how a proposed intervention might affect a segment's risk would be a valuable feature.
Having historical intervention data available would obviously help - if a user wanted to know what impact installing a stop sign would be, we could search for similarly profiled segments that were treated with the same intervention, and use that segment's change in risk as a guide (controlling for other variables as necessary).
In the absence of historical interventions, I wonder if we could have a "second best" form of estimation, whereby we find a similar segment to the one in question that already has the feature being considered (but date of installation unknown) and compare the risk between the two. This feels a bit blunt though, any other ways we could approach this prior to having a history of interventions?