Biological systems are inherently variable. Variability in the spatial and temporal distribution of species within an ecosystem means that it is often difficult to separate natural variability in a measured parameter (often referred to as ‘noise’) from any causative effect. For example, it is difficult to ascertain whether a difference in the abundance of species observed in two populations is a result of chance, or a result of some significant underlying difference between the two populations. An important aspect in the design of any monitoring survey programme is to minimise the degree to which the natural variability within the measured data affects the statistical analysis and interpretation of the data.
There is the potential to make two kinds of error in the interpretation of statistical models. Type I errors (indicated by α) are made when the null hypothesis (H0) is rejected when it is in fact true. Type I errors are also referred to as the significance level, or p-value. Type II errors (indicated by β) occur when H0 is accepted when it is false (Table 2.1). Statistical Power is a measure of confidence that a statistical analysis will give us the “true” answer by limiting the risk of committing a Type II error.
Power is therefore simply 1 – β. It can be carried out a priori, using information gained from a pilot study or the literature to inform on the number of samples required to allow for robust statistical analyses (e.g. pre and post construction studies), or post hoc, to assess whether results from analysis are valid (Quinn and Keough, 2004).
Table 2.1. Summary of Type I and Type II errors Truth for population
H0 is true Accept H0 Reject H0
Correct (True positive)
Type I (False positive, α)
H0 is false
Type II (False negative, β)
Correct (True negative)
The two types of error are inversely relational and an increasing effort to reduce β increases the risk of encountering Type 1 error. It is therefore common practice for a compromise level for Power to be set at 0.8 (Crawley, 2011).
The formal representation of Power analysis is: Eqn. 1.
√ Where:
σ– Standard deviation. A measure of deviation within dataset. The greater the standard deviation in a data set, the greater the degree of variation of data
