Time Series Analysis of Bod Data Using the Gibbs Sampler

In this paper, we analyse the biochemical oxygen demand (BOD) data collected over 2 years from McDowell Creek, Charlotte, NC, by fitting a trigonometric time series model to the data via a Kalman filter and by using the Gibbs sampler. Some graphical di agnostic tools are applied to monitor the convergence of the algorithm. The adequacy of the model for one-step ahead prediction and simultaneous prediction problems using the extended Gibbs sampler is studied. The problem of detection of outliers is also discussed and the BOD data is analysed for outliers.