|
Abstract:
|
Often, there is interest in monitoring health within a study region where data are available for a number of subregions. One way to carry out monitoring is to maintain separate cumulative sum charts over time for each region. A drawback of this approach is that it does not account for the possibility of clusters occurring on larger spatial scales. In this paper, I describe how monitoring may be carried out for neighborhoods constructed around each subregion. Separate charts may be kept for each subregion and its surrounding neighborhood. However, adjusting cusum thresholds for the number of subregions is conservative, as nearby subregions will have correlated charts. Here, these correlations are accounted for; an adjustment for the number of effectively independent charts is made using the theory of smoothed Gaussian random fields, and the approach is evaluated.
|
