The among-site precision of metrics and indices may be assessed among sites
within the same reference class, thus limiting the confounding effects of environmental variability on the evaluation. For both O/E indices and MMIs, the ratio of
observed-to-expected metric values at reference sites should differ negligibly from
one, and thus the standard deviation (SD) for this ratio should be nearly equal to its
coefficient of variation (CV). The distribution of observed-to-expected metric or
index values provides a graphical illustration of both accuracy and precision
(Fig. 5) [72].
Fig. 4 Box plots of simulated data illustrating the method of Barbour et al. [133] for evaluating
the discriminatory power of metrics and indices. Boxes represent 25th and 75th percentiles;
whiskers represent non-outlier maximum and minimum values. The metric is expected to decrease
with impairment. Top left: discriminatory power ¼ 0 (lowest), as the reference and impaired site
interquartile ranges (IQRs) overlap and include both medians. Top right: discriminatory power ¼ 1
because the IQR overlap includes only one median value. Bottom left: discriminatory power ¼ 2
because the overlap does not involve either median. Bottom right: discriminatory power ¼ 3
(highest) because the IQRs do not overlap
Principles for the Development of Contemporary Bioassessment Indices for. . .
251
within the same reference class, thus limiting the confounding effects of environmental variability on the evaluation. For both O/E indices and MMIs, the ratio of
observed-to-expected metric values at reference sites should differ negligibly from
one, and thus the standard deviation (SD) for this ratio should be nearly equal to its
coefficient of variation (CV). The distribution of observed-to-expected metric or
index values provides a graphical illustration of both accuracy and precision
(Fig. 5) [72].
Fig. 4 Box plots of simulated data illustrating the method of Barbour et al. [133] for evaluating
the discriminatory power of metrics and indices. Boxes represent 25th and 75th percentiles;
whiskers represent non-outlier maximum and minimum values. The metric is expected to decrease
with impairment. Top left: discriminatory power ¼ 0 (lowest), as the reference and impaired site
interquartile ranges (IQRs) overlap and include both medians. Top right: discriminatory power ¼ 1
because the IQR overlap includes only one median value. Bottom left: discriminatory power ¼ 2
because the overlap does not involve either median. Bottom right: discriminatory power ¼ 3
(highest) because the IQRs do not overlap
Principles for the Development of Contemporary Bioassessment Indices for. . .
251