The analysis of biodiversity data is a highly context-dependent process that is defined by the underlying goals and objectives of monitoring, the limitations on any potential inferences that are imposed by the choice of indicators and sampling design, and the availability of relevant theory and natural history knowledge to help interpret the ecological significance of any results.

The four main types of data analysis that are relevant to monitoring relate to problems of: describing the biodiversity to be used in analyses; detecting change with respect to performance standards (i.e. as part of effectiveness monitoring); understanding the reasons behind observed changes (i.e. as part of validation monitoring); and predicting patterns of biodiversity across space and time.

Decisions about how biodiversity data are summarized and manipulated prior to any modelling work can have a significant influence on our perception of management impacts. Important considerations for improving the quality of sample data include: analysing the response patterns of individual species, as well as entire assemblages; employing caution when using estimates of species richness and diversity indices; testing the sensitivity of observed patterns to differences in spatial scale; understanding species-mediated ecological functions; assessing species-specific conservation priorities; and using biotic integrity indices to improve dissemination of monitoring results.

Confidence intervals provide a more defensible and practical approach to assessing management compliance against performance standards than significance tests of a null hypothesis of no change in indicator values.

Statistical modelling provides a means of formalizing our understanding of the environmental impact of human activities and identifying the extent to which our predictions are supported by sample data.

Information-theoretic and Bayesian approaches to the analysis of monitoring data have distinct advantages over more traditional methods that are based on null hypothesis testing (NHT) because they are explicit about uncertainty and allow the evaluation of a range of alternatives. 258

While statistics provide an essential and powerful tool for biodiversity monitoring, they are only one piece of the puzzle and their usefulness can be readily compromised by uncertainty in sample data or a lack of relevant ecological knowledge about the study system.

To be most useful the analysis of biodiversity data needs to be conducted in the context of trade-offs between multiple management objectives.