The use of chi-square maps in the analysis of census data

This paper addresses itself to one of the problems inherent in any spatial analysis of census data, namely the variable size of the enumeration units. This variation renders some proportions unstable, due to the small numbers involved. Attempts to overcome this, using visual methods, can be unsatisfactory, whilst doing nothing to aid statistical analysis. Chi-square maps are suggested as a satisfactory alternative in both contexts, and data for Reading, Berkshire are displayed to illustrate this. These maps suggest that some inferences, based upon indicators derived from proportions, may be unsound.