Eutrophication due to increased input of nitrogen in the Neuse River Estuary in eastern North Carolina, USA, was considered the primary cause of large scale fishkills in the late 1990s. The North Carolina General Assembly established laws to protect the estuary, including a requirement of reducing nutrient (particularly, nitrogen) input to the estuary. Because eutrophication in North Carolina is measured by the concentration of chlorophyll a, assessing the success of the nutrient reduction program relies on the demonstration of a reduction in chlorophyll a concentration in the estuary. Three institutions have water quality monitoring programs including chlorophyll a in the variables they measure: NC Division of Water Quality (DWQ), University of North Carolina Institute of Marine Sciences (IMS), and Weyerhaeuser Corp. (WEY). Because the estuary is large and chlorophyll a concentrations vary spatially, methods used in sampling and measuring can affect the reported chlorophyll a concentrations. To demonstrate the success of the nitrogen reduction program, we need to compare the chlorophyll a concentrations measured before and after the implementation of the program to the concentrations before. But which series of data should we use? To answer this question, we need to compare the reported chlorophyll a concentrations from the three institutions and determine whether they are different. If they are the same, we may want to combine the three sources of data to increase the power of any statistical test we will use. If they are different, we need to describe the nature of the difference and decide how to best use them to describe the effect of the nitrogen reduction program.
For this problem, we are to compare the chlorophyll a concentrations from the three institutions and discuss the differences among them:
â€¢ Exploratory data analysis â€“ a summary of data distributions and potential problems with the data.
â€¢ A decision on whether a transformation is necessary. In general, we use log-transformation for environmental concentration variables.
â€¢ ANOVA to test whether the mean (or median, if log-transformed) varies by institution.
â€¢ Present the estimated differences (and interpret the differences in the original concentration scale)
â€¢ A short discussion on other factors that may affect the result of this comparison.