The first, and most important factor in developing an appropriate statistical design, is knowing the objective(s) you’re asking (your scientific hypotheses) and hence the type of question(s) you are seeking answers to. The objectives will have direct, and sometimes obvious, implications for design. As an example, if the objective is to estimate the effect of implementing a management zone (e.g. a no-take reserve) - a casual type of research question - then at a minimum samples will have to be taken from outside the reserve (a control) as well as within it. We think it useful to consider the following list of probing questions before starting the design process.

  1. What is the primary research question? Is a comparison between areas of interest required (e.g. impacted/not-impacted)?
  2. What is the appropriate metric to measure (and to subsequently analyse)? Often measurements will be taken on species (e.g. biomass, size, presence-absence, and/or abundance), but analysed as a different quantity (e.g. a diversity index).
  3. What are the primary sources of potential difference, if any? This will depend on the research question (e.g. impacted/not-impacted areas) but may also include extraneous variables such as environmental conditions and human impact.
  4. Could locations with different ‘treatments’ also differ in other important variables, so that there is potential for confounding? An example is whether the habitats within and outside no-take reserves are different.
  5. What resources are available for conducting the survey? Is there a particular type of sampling platform that should be used (see the remaining chapters)? What previously-collected information is available to aid the current survey (e.g. bathymetry or back-scatter data)? How can we use the previously collected information? Does the previously collected information make one sampling platform or unit of measurement a better candidate than another? How many samples can be taken? This last question also directly affects the power of the survey to detect differences in contrasts of interest. See Section 2.5 for more discussion on power.