Search This Blog

Friday, 15 October 2010

Populating the grid squares

I am reasonably happy with the results shown above. This is populating my grid squares with human residential population data obtained from the Office of National Statistics 2001 census. I have used a similar technique to that which I discussed with IMD scores in my last post with a few variations. Firstly I have used population data from the smallest census unit, the Output Area(OA). There are 24140 OAs in London which is a remarkably similar number to the 26116 grid squares, but OA are based on a residential poulation of about 300 people and therefore are of vastly varying sizes. I have experimented using average numbers of people but this gives a highly inaccurate result because there is a bias to OAs with large spatial areas as the grid squares falling within them all having the value of the OA which instead of being counted once will be counted many times. So instead I have gone for average density (remembering that to go via the average number of people and average area rather than averaging the individual densities) to counter-act the area bias. When added all togeather there is a resulting error of about 10% too high a population and about 30% too high an area resulting an average density in London of about 37 people per hectare rather than about 47 which the actual data supply a resultant error rate of about 20%.

This is acceptable to me because (a) the 2001 data is dated (b) the relationship between residential population and crime and disorder I have already argued on this blog is not straight forward.

The naive smoothing is there (this is what causes the error) but to me it is intuitively more of an advantage than a disadvantage.

What is possible from the Census Data is use different age groups in the population. These are what the following maps represent.

0 to 15 years old inclusive


16-29 years old inclusive


30-59 years old inclusive


60 years old and older

No comments:

Post a Comment