Figure 4-9(B) illustrates that, under the conditions specified, it would take between 5 and 20 years (equal to 1 

 to 5 sampling cycles) for the indicator to detect the specified magnitude of change (2% per year if the regional 

 median = 50 points) with a power of 0.8, given a coherent variance across years of between and 1 00 points^ 

 of indicator score. Figure 4-9(C) shows that, under the specified conditions, it would take between 5 and 13 

 years to detect various changes in indicator scores, representing 1 to 4% change per year in a regional 

 population median score of 50 points. Figure 4-9(D) shows that, under the specified conditions, it would take 

 between 7 and 9 years to detect the specified trend, depending on the number of sites visited per year. This 

 series of figures points out that the capability of the indicator to detect trend is affected most by the magnitude 

 of coherent across-year variance and by the desired magnitude of change that the monitoring program is 

 expected to detect, and is affected to a lesser degree by the sample size. These analyses need to be 

 repeated once a more robust estimate of coherent variance across years is obtained from several years' 

 worth of data to determine which of the scenarios presented in Figure 4-9 is the most realistic . If the coherent 

 across-year variance in the indicator score is relatively small (< 10 points^) the indicator should meet the 

 performance criteria for both status and trend established for EMAP-related monitoring frameworks. 



Table 4-18. Statistical power capabilities 



Power to discriminate among classes of ecological condition 



Proposed monitoring framework: a minimum of 3, and preferably 4 classes of impairment in 

 condition are desired: 



• Fore ef a/. (1996): Analysis of similar multimetric indicator suggests 5-6 classes of 

 condition can be distinguished at a = 0.05 and p = 0.2 



• Similar approach, using sites with repeat visits and/or resampling methods such as 

 bootstrap procedures, is potentially feasible with indicator 



Performance criteria for proposed monitoring framework: 



• 90% confidence interval should be < 1 0% of the estimated proportion of a resource 

 that is at or below a threshold value designating impairment 



Power to detect trend in condition 



Performance criteria for proposed monitoring framework: 



• Magnitude of trend: 2% per year change in regional population median indicator 

 score (= 20% change over a 10-year period) 



• Sample size=50 to 200 sites monitored in region per year 



• Probability of false positive (a) = 0.1 



• Probability of false negative (P) = 0.2 (Power = 0.8) 



Summary 



Results from a previous study imply that 3 or 4 classes of condition can be distinguished over the potential 

 range of indicator scores. Preliminary analyses indicate performance criteria for both status and trend detection 

 can be met if across-year variance is relatively small compared to within-year variance. These analyses 

 must be repeated after more robust estimates of coherent variability among sites are obtained from several 

 years of data collection, and after the responsiveness of the indicator (Guideline 12) has been adequately 

 established. Power curves were used to demonstrate the effects of alternative monitoring requirements, 

 especially the importance of coherent across-year variance and desired magnitude of change. 



4-37 



