Search This Blog

Monday, 21 December 2009

Starting to bring it together

Starting from today I am going to post data relating to street robberies in a London borough. It is easy to keep the borough unidentified in graph form but less easy in map form. I will see how I do. The data is deliberately historic. Its source is the initial classification of robbery allegations recorded on the Metropolitan Police crime recording information system CRIS. In recent posts I have touched on time and yesterday I introduced demographics and geodemographics. This graph shows the demographic of age against time. It clearly shows that robberies which have victims who are under 20 years of age occur at different times to those people of older ages. The times suggests school going home time as the vulnerable time for the lower age group. This analysis is reinforced by a graph of Saturday and Sunday robberies.
Now I have to map these robberies out to see what geographical relationship, if any, they have with schools.........


  1. Interesting visualisation, there are obviously some characteristic trends coming out - however using pseudo-3D Excel graphs is flawed at best. For example, on the top graph, what are the values for under-20s between midnight and 6am? They might be high but obscured by the 20s-30s data. Obviously just a quirk of the 3D graph in thie case, but for more controversial data, cynical people might think there's potential for deliberate bias, or hiding undesirable data to reinforce a hypothesis.

    Using small multiples would not be hiding any values or giving undue prominence to other data and therefore would be "fairer". Do you know Tufte's seminal work "The Visual Display of Quantitative Information", which goes into great detail about this kind of thing?

    Of course, it takes a lot more time to set up small multiples than just pile everything into a pseudo-3D graph!

    By the way I think the borough is Westminster.

  2. Small multiples are all well and good, but why not just go for a standard 2D line graph? That should show everything required without having to wrestle with the vagueness of a 3D perspective.