Some Advanced Topics

1. How do I find an optimal classification to apply to my data?

2. Labelling

3. How can I draw maps based on Leeds GAS output files


1. How do I find an optimal classification to apply to my data?

There is no uniquely optimal classification that will work on any data. You need to find a classification that is best for a given application. Much depends on how you measure performance - this is subjective and context dependent.


2. Labelling

Cluster labelling is the task of deciding what to "call" or "label" the clusters. This is a minefield of complexity both methodological and conceptual with wide ranging possible implications and ramifications. The aim is simple enough - find a short descriptive prose that captures the principal characteristics or distinguishing features or unique aspects of the cluster. This is hard in practice because it is subjective, and arbitrary, and partial; it will often highlight very distinctive but clearly minority features; the words used can be misleading, hurtful, as well as incorrect; it propagates ecological fallacies; it may violate Data Protection principles if it is subsequently used to deny the inhabitants of specific clusters services or facilities, but it is also essential.

There are two possible solutions:

Use no labels and apply the cluster codes as a numeric summary index or invent your own.


3. How can I draw maps based on Leeds GAS output files?

You need to import the GAS results file back onto your local machine. You then need to acquire digital boundaries for the geography you have used. MIDAS and UKBORDERS will supply these to you. LGAS can provide your results by 1991 census ED and you can use these as a basis for mapping at either ED or ward or high levels of census geography.


Return to: Help Topics  LGAS homepage