How to Lie with Statistics

On Thursday, HalifaxExaminer published this announcement of a Government meeting:

Crosswalk safety advisory committee (10am, City Hall) — there’s been a welcome decline in the number of pedestrians struck:

I don't know if Tim is just pulling our leg, but nothing could be further from the truth. As I have said before on this subject, the police's insistence on viewing the complex world one variable at a time can lead to false conclusions and, most importantly, neglects important information. A simple example:

Rather than looking year-by-year, here is a graph of the complete 3 1/2 years, showing that 2015 has barely nudged off the moving average of 17.1 incidents per month (The lowest moving average, 14.7, occurred in August, 2013):

This can be confirmed by averaging the data by month and comparing it to the first seven months of 2015:
The downward trend of 2015 seems to occur in the first half of every year.

Better yet is a summary by season, which does provide a clue as to what's happening:

You don't suppose that pedestrian collisions are connected to lousy snow clearing? Sounds pretty far fetched......

I promise you that the number of collisions will soon rebound to the 15-20 range, peaking in the winter months.  As a wheelchair user, I take this personally because I'm not very agile in avoiding hurtling autos.  

The chart of Vehicle-Pedestrian Collisions is entirely inconclusive as to any trend and further demonstrates HRM Police’s confusion of ‘counting’ with ‘analysis’. The stubborn reliance on considering one variable at a time means that important questions go unanswered.

• Which intersections have accidents during peak commuter times?
• Are there intersections that are dangerous for under 25-year-olds on weekends?
• Do lighting conditions make a difference?
• How about the pattern of paint?

Intersections are complex, and the problem will persist until the complexities are accounted for.  Arithmetic just isn”t enough.