Crime Data Not Showing Correctly

[AmandaWilsonKC]

It seems that the crime data points on the map don't seem to match up with the numbers in the drop down report. For example, in the financial district, the drop down report says there are 44 homicides in the previous year but the map data shows 7 homicides and only a portion of those were last year.

[AmandaWilsonKC]

Another place for pulling crime data would be the County's GIS. Need to compare data from the County's GIS and the City's open data but i wonder if our app could easily help compare the two??

http://arcgisweb.jacksongov.org/arcgis/rest/services/COMBAT/ViolentCrimePoints/MapServer  Part I Violent Crime (County-wide, dates depend on agency – KCPD data through last week)

http://maps.jacksongov.org/combat/  COMBAT’s web mapping application – it has the Part I Violent Crime data

[moldybeats]

Seems like these Jackson County COMBAT URLs have moved:

https://jcgis.jacksongov.org/combat/combat.htm

https://jcgis.jacksongov.org/arcgis/rest/services/COMBAT/ViolentCrimePoints/MapServer/0

[buzwells]

Since I am already working on the crime tab, I can take a look at this issue. After I close out the https://github.com/codeforkansascity/Neighborhood-Dashboard/issues/142, I will do a deeper dive into the data to try to reconcile our own views internally. Then I will try to compare our data to data from other sites.

[buzwells]

I added some browser console logging today to help illuminate some of the issues. What I found was that map pins often represent multiple offenses, but that is not indicated anywhere on the pin. In a little testing with assaults in one neighborhood, the grouped totals (crime report) showed 12 for 2018, and a query I ran also returned 12 assaults. But only 4 pins dropped on the map due to location overlap. Particularly with crimes, where the locations are simpler (just the intersection), this is going to be true. I only tested with one type of crime (assault), but I'm guessing the same would be true if more than one type of crime occurred. This even challenges the legend a bit (though we could add a pattern for multiple types of crimes).

Here's a screenshot with the above-referenced data. You can see 12 offenses, but only four addresses.

How do we want to handle this in terms of both the pin colors and also the popup content? I'm assuming we will want to list all the crimes in the popup, but that could get lengthy.

[buzwells]

Looking more closely at the data, it appears that we are counting each record in the dataset as a crime. But it looks as though there can be multiple rows for a single crime, all linked together by a "report no" field. Each separate row appears to relate to a different person involved in the crime (e.g., suspects, arrestees and victims, identified by the "involvement" field). If this is correct - and I certainly need to do a little more digging on this - then we probably only want to count once for the whole incident.

[buzwells]

I pushed and merged some improvements to the display of crime data:

• The crime report now counts each incident only once. Previously, we were over-reporting incidents by counting each involved person as a separate crime.
• Merge incidents at the same intersection into a single marker.
• Display an incident count for each marked location. I think this is an improvement, but there is still more fine-tuning to do. For instance:
• For some neighborhoods, counts by marker are off by one or two from the counts in the crime report (when using last year in the Filters). I believe this is because the report is driven by a query that assumes everything in the database for a particular year relates to that year (no date conditions specified). Adding a date condition should fix the problem, but will test to see.
• Our legends assume each location has only a single crime category. Because locations can now have multiple crimes, they can also have multiple categories of crime. Will need a design to handle that.

[buzwells]

Another additional issue: some filter checkboxes are not sending down the right codes (they appear broken), which results in no pins being dropped even when the violations exist.

[buzwells]

There were 4 crime filters that were not working properly because they were not associated with any codes behind the scenes: Sex Offenses (both), Counterfeiting and Domestic Nonviolent. I've made the necessary fixes, including changing the label of "Counterfeiting" to "Forgery/Counterfeiting" to reflect the underlying NIBRS documentation. I will be pushing and deploying the changes shortly.

[buzwells]

Fix has been deployed. Closing this large issue. Will open smaller issues for minor remaining improvements.