Trend Evaluation Using a Rotating Panel Design and Multiple Imputation
Don L. Stevens
Statistics, Oregon State University, Corvallis, OR, United States

The evaluation of “status and trend” is a common objective for environmental monitoring. Most often, the focus of the monitoring is not a single site, but a political or geographical region. Thus, interest is in a population-level description of trend or regional trend. The traditional statistical concept of regional trend is trend in population parameters, where the usual population parameter of interest is the mean. Trend is described by sampling the population and estimating parameters at distinct points in time. The resulting estimates are then analyzed for trend, e.g., with time series or regression methods, or tested for significance. However, environmental populations can exhibit substantial change and still leave population parameters invariant, so a more general concept of trend is useful.

The distribution of site-specific trends provides a very general context for examining regional trend. The site trends are calculated from repeated observations at the sites, using whatever trend descriptor is suitable (time series, regression, complex models with ancillary variables). Regional trend is then summarized by the distribution function and its properties (mean, median, percentiles).

Rotating panel designs have been promoted as an effective sampling design for the dual objectives of status and trend. This talk illustrates the use of multiple imputation with rotating panel design to estimate the distribution function of trend using data from long-term monitoring of Coho salmon on the Oregon Coast.

Keywords: Rotating panel; Multiple imputation; Trend evaluation

Biography: Dr. Stevens has over 30 years of experience in applying statistics to problems arising in the physical and biological sciences. For over ten years, he was engaged in developing the statistical sampling theory supporting the USEPA's Environmental Monitoring and Assessment Program spatially balanced probability sampling, and simultaneously applying that theory to designing samples of a variety of aquatic resources. He has worked with the Oregon Department of Fish and Wildlife since 1995 on monnitoring salmn populatiions.