2 ms·
Nice walkthrough of the process of going from broad to narrow and then trying to explore the odd outliers. At this stage -- the point where you've identified a
by danso 11y ago
Nice walkthrough of the process of going from broad to narrow and then trying to explore the odd outliers. At this stage -- the point where you've identified a noticeable spike in arrests on early Wednesday's evenings -- it's worth trying to ask the SFPD why this discrepancy exists -- if it's something related to the reality of arrests (such as drug bust patrols stepping up on early evenings) or even what the Date/Time field represents -- is it the time of the initial report, or when the booking has been made? I've looked at the SFPD data on various occasions and have never been able to figure it out (based on looking at the data alone) and I wouldn't be surprised if it varied depending on the crime and disposition.
One thing I noticed on a recent inspection (actually, one of my students pointed out) was that the arrest rate is lower in 2015 for certain categories of crime, compared to 2014...and one possible reason for that discrepancy is because the dataset is updated as the disposition of the incident changes...i.e. when an arrest is finally made, a month(s) later, then the record is updated in the dataset.
The SF crime data is certainly granular and lends itself to analysis (certainly more than what's released by the NYPD)...but it lacks the specificity that can be found in Chicago and Dallas crimesets, which, IIRC, have separate fields for recording the time of the report and the time of last update to that report, among many other fields (though all the various departments have their own flaws and inconsistencies).