Simulation of traditional and optimal sampling schedule designs for digestion and passage models

Abstract

The determinants of the Fisher information matrix and accuracy of parameter estimation were evaluated for a combined design and four even, four geometric and three D-optimal sampling schedule designs using either a digestion or passage model. Initially, three sets of parameters were used to calculate three D-optimal sampling schedule designs and then these optimal designs were tested with other parameters. The combined design was a combination of the sampling schedules derived from the initial D-optimal designs. The D-optimal designs were, in general, not robust when parameters were varied and were more sensitive to changes in the passage model parameter values. The combined design was not the best scheduling design in all cases but was the best design on the average. The even sampling schedules were the worst designs for the digestion model but were the second best designs for the passage model. The longer time geometric designs performed as consistently but not as well as the combined design for both models and as well as the even design for the passage model. The results from this study suggest that using a design that may not be the best design for a particular set of experimental conditions will be the most robust design over a range of conditions.