You are here:
Multimetric indices: How many metrics?
VAN SICKLE, J. Multimetric indices: How many metrics? Presented at 58th Annual Meeting of the North American Benthological Society, Santa Fe, NM, June 06 - 11, 2010.
Multimetric indices (MMI’s) often include 5 to 15 metrics, each representing a different attribute of assemblage condition, such as species diversity, tolerant taxa, and nonnative taxa.
Multimetric indices (MMI’s) often include 5 to 15 metrics, each representing a different attribute of assemblage condition, such as species diversity, tolerant taxa, and nonnative taxa. Is there an optimal number of metrics for MMIs? To explore this question, I created 1000 9-metric MMI’s by randomly choosing 1 metric representing each of 9 attributes of condition, from 113 candidate metrics estimated for 161 aquatic vertebrate assemblages sampled during the 2000-2004 EPA-EMAP survey of Western USA streams. Over all 1000 cases, performance statistics for MMIs declined when metrics were removed at random from the 9-metric MMIs. Performance increased slightly, but at a declining rate, as metrics were added to the 9-metric MMIs. However, among the top 10% of 9-metric MMI’s, performance stayed flat or declined as metrics were added. Similar results were seen for 5 other MMI-development data sets. Results suggest that best-performing MMI’s can be built using a single metric for each attribute of assemblage condition, and that performance is unlikely to be improved by adding more metrics.