Analysing correlated count data from field trials

Field experiments are often affected by both spatial and temporal (i.e. repeated measures) correlation. In order to obtain an analysis that is scientifically valid it is important to recognise the underlying error structure and analyse the data accordingly. We will discuss the analysis of count data which is spatially and temporally correlated, and illustrate the difference between an independent error structure model and a marginal Quasi-Likelihood model which attempts to account for the correlation present in the data. We shall then show the possible impact of inefficient analysis techniques on the subsequent economic decisions.