Regions for Health collaboration methods

Clustering itself; K-means clustering (non-hierarchical algorithm) Based on FASTCLUS in SAS - Four iterations using the standardized data

Number of clusters depends on number of regions (Canada started with 130, now 105). Number of regions divided by seven (will give around 15 regions per cluster). Region which appears in only one cluster is defined as an outlier. RMS should be as low as possible. Radius is the distance between regions and should be as low as possible.

Perform step-wise discriminant analysis