Analysis of time-space clustering in veterinary epidemiology

Techniques useful for investigating time-space interaction in the clustering of events in veterinary epidemiology--the Mantel test, Barton's method, nearest-neighbour test and Knox's test--are described. The use of these techniques is demonstrated by the analysis of a data set (containing blowfly catches on a commercial sheep property located in Tasmania, Australia, during the period December 1997-May 1998) in which clustering was apparent. Significant (P<0.05) clustering of blowfly catches in the data set was only detected by Knox's test. The use of Knox's test provided an insight into the possible clustering of blowfly catches over short (< or = 3 km) spatial and temporal (< or = 1 month) distances. Results demonstrate that several techniques should be used when attempting to identify and describe whether events are clustered in time and space (to maximise the power of the analysis). Also, it is important to consider the spatial and temporal models implicit in techniques chosen when interpreting results of analysis.