Planning Your Monitoring Project
Use this analysis to:
- Balance cost and sensitivity
- Interpret results
- Balance sites vs replicates
- Request realistic budget
- Discuss needs with stakeholders
- Ensure the study is useful
Here is an example of statistical power analysis. This statistical tool is an essential aid to help prioritize the level of sampling at different sites. In this particular study, a Hydroelectric facility needed to assess impacts of a change in flow management on stranding (and killing) fish in small pools along the river bank. A poorly designed study would produce a "no significant difference" result even if one flow-method resulted in many more fish being stranded on the river bank!
This figure is the result of some statistical data manipulation from a prior similar study of fish stranding and shows the estimated number of samples required (x-axis) to statistically describe the minimum detectable difference in fish stranding. The baseline (dark blue line) assumes that with current flow management protocols an average of 2 fish are stranded per flow cycle. If only two samples were collected, the new flow flow cycle would require 8 fish to be stranded (7.5) before statistical test would determine the change to be significant.
We found that with 9 sample locations, a change in stranding to 4 fish (2 more than the current management strategy) would be statistically significant (see teal line), and that additional sites would not improve the power of the monitoring program significantly (light blue line). We also found that five sites could be used if the stakeholders were satisfied with not being able to detect a change of 4 fish (red line). That is, if they are OK with killing five fish per flow cycle in the study area, then they could use five samples rather than nine. Also, note that nearly 20 samples would be required to detect very subtle change to three stranded fish... thus arguments for 12 samples were not worth the cost.
