![[logo] US EPA](http://www.epa.gov/epafiles/images/logo_epaseal.gif){kind=logo}
Status Assessment
Description: When the entire population includes too many items to measure them all, we must sample. If we sample randomly, we can apply our conclusions to the entire population using statistical inference.
Don't get confused by the word "sample," as its meaning changes in different contexts. In field biology we talk about "samples" of invertebrates, 1 for each stream site. At a higher level, when we talk about sampling design, we mean the stream sites within a region. In this case, the population is all possible stream sites in the region, and the sample is the sites we actually visit.
Don't get confused by the word "random" either. Random selection implies an explicit probabilistic model, not selection based on whimsy or haphazard choice.
Simple example: Surveys published in the newspaper, such as opinion polls, are based on a sample of people. The validity of any poll depends on whether the population of interest (e.g., registered voters) was randomly sampled. Often, answers to survey questions are accompanied by an estimate of error. That error is a measure of uncertainty related to the answers of people who were not surveyed.
MAIA example: A multimetric fish index was used to summarize the condition of the fish assemblage within the MAIA region (Stoddard, pers. comm. and McCormick, et al. [in review]). Because the initial sample was random, they could apply their estimates to the entire population of streams in the region, including streams that were not sampled. Alternatively, they could compare stream populations within different ecoregions and determine that streams in the Ridge and Blue Ridge ecoregions were in better shape than those in the North-Central Appalachians.
Figure
Figure. Comparison of fish condition in 2 geographic areas. Green indicates good condition, yellow indicates moderate, red indicates poor and gray means insufficient data.