Data on cyanobacteria (blue-green algae) are gene:rally collected on a
reactive basis, frequently in response to bloom events. Such data pre
sents a biased and incomplete snapshot of water quality. This paper lo
oks at two typical data sets for UK waters showing that while statisti
cs may be used to describe the data they are of limited use in forecas
ting. Suggestions of appropriate tests for small and sparse data sets
are made.