I'm continuing to think about the "contest" from Open Data Discourse. I managed to deal with the sunset data and also processed the data so I can geocode in QGIS. First result: 4PM to 7PM, Monday through Friday. Not unexpectedly, nearly all of the accidents are on the major commuting routes. It thins out if you look at the same period but for after sunset:
If you look at the after work commuting hours after sunset, you see that most accidents are on Mass Ave and the major routes leading into, and in, Inman Square but the density is far less.
The short answer, so far: riding at night, on major commuting routes is probably a less than safe time and place to ride but accidents happen during daylight.
Sunset times were derived from here but I used only 2010 data, which is within 1 minute for times for other years. I used this site for timing of daylight savings time for the 4 years.
The table is here, if you want to look at it directly.
See the first post on mapping bike accidents here.
Two thoughts about the data. 1) these are reported accidents only. I don't have any idea of what accidents are not reported and there are certainly no near misses in this dataset and 2) I noticed a significant (so significant that I remembered it from 3 years ago) accident at the corner of Mass Ave and Vasser. This accident was not in the data set.