values about the mean and the more difficult it is to measure statistical differences between populations (e.g. pre- and post-development, between geographic locations, etc).
es – Effect size.
α – Required significance level, or p-value. The desired p level is to be set at 0.05 (or 95% probability).
n – The sample size
Any one term from Eqn. 1 can be solved when all other terms are known. 2.1 Standard Deviation
Standard deviations for epibenthic and benthic community mean values for this power analysis have been calculated from the 2010/2011 Zonal Characterisation Survey data (MESL 2011). Data with higher standard deviations (δ) have a greater degree of variability. These data require a higher degree of sampling effort (i.e. a higher number of samples, n) to detect a significant effect of a given size than data with low variation values.
2.2 Effect Size Effect size is the level of change we are able to detect. It is calculated thus: Eqn. 2 (Coe, 2002)
and can be summarised as standardised mean difference.
In this study we are attempting to understand if a suitable number of samples have been taken as part of the Zonal Characterisation Survey to correctly detect a response to development of East Anglia THREE and FOUR and the Cable corridor in subsequent post-construction surveys. The effect size required that is representative of a true shift in the benthic and epibenthic communities however, is unknown. This is a common issue with many Power Analyses and to satisfactorily deal with this Cohen (1988) proposed the use of effect sizes of 0.8, 0.5 and 0.2 to represent a high, medium and small effect size. These effect sizes reflect degrees of change based on standard deviations (e.g. 0.2 = two standard deviations from the mean of survey A). This study has also investigated the use of a range of es (see Section 4.2).
Power Analysis calculations were based on calculations of Shannon Diversity and Simpson’s Index at each survey site for benthic and epibenthic fauna. For ease of interpretation, this study will present es as percentage change. Although this is not common practice for power analysis it has been possible to convert es to an Estimated Detectable Percentage Change (EDPC) by partially solving Eqn. 2. Data from the Zonal Characterisation Survey were assessed and the standard deviation within
