Re: [Gnumed-devel] Interesting statistic on doctor demographic density in BC CA

[Jim Busser]

On 2011-10-11, at 1:42 PM, Dave Cramer wrote:

>> I find that there are about 2100 distinct streets or about 6 doctors per 
>> street.

select count (distinct (id_street, number, subunit)) from dem.address ;

 count 
-------
  5081

Among 12,200 doctors this makes an average of 2.5 doctors per building subunit

(mind you, they may have variably spelt their subunits and might collapse to a 
higher density).

The density per building having the same number is

        12,200 / 3180

or 3.8 per building.


Despite that my importer script (to clean, map and import the source data) is 
already 2200 lines of SQL and I did not even yet create the org_units nor the 
links to the org addresses (to say nothing of the comm channels) it would not 
really be possible to figure out such data from a mix of

        post office boxes
        apartments, street numbers, street names
        street names, street numbers, apartments

which is one typically inherits with non-granular, non-granular 'addresses' of 
the sort that are typically available.

I guess we can either require to do this sort of work up front (by getting your 
data into a granular form to start with) or post hoc (working with non-granular 
and non-normalized data). Which latter means it usually won't happen. Otr, when 
it does, that it will have limited accuracy.

So, a thanks to Karsten for the discipline on the storage requirement!

-- Jim