Time-Dependent Models With Correlated Obse Rvations
Anthony C. Atkinson, Alexander N. Donev, Randall D. Tobias · 2007
Abstract Observations that occur as a time series may not be independent. In this chapter we see how the methods of optimum design have to be adapted if there is indeed correlation between observations at different time points. The nonlinear examples of Chapter 17 illustrate the general point. In Example 17.3 with two consecutive first-order reactions, the concentrations of chemicals evolved over time. On the assumption that the errors were independent we found the D-optimum design for the two parameters, which consisted of taking observations at just two time points.