3.4 Spatial clustering or variation in risk?

Distinction between methods described in Sections 3.2 and 3.3.

In both cases,

• •

  Null hypothesis: cases occur independently AND with constant risk

• •

  Alternative hypothesis:

  1. 1.

     spatial clustering: stochastic dependence

  2. 2.

     spatial variation: non-constant risk

Distinction between stochastic dependence and non-constant risk is not sustainable empirically (and not always theoretically, without additional assumptions) (Bartlett, 1964).

A pragmatic strategy for problems of this kind, is to

• •

  try first to explain spatial variation in risk by adjusting for covariate effects before using stochastic models;

• •

  then use stochastic methods to analyse residual spatial effects

• •

  interpret small-scale residual effects as clustering, large-scale residual effects as variation in risk?

• •

  a scientific explanation always trumps a purely statistical one.