Profile-Based D-Optimality for Designing Experiments in Nonlinear Situations

On behalf of the Department of Mathematics and Statistics of the College of Arts and Sciences, you are cordially invited on a seminar to be conducted by Dr. Hana Sulieman, AUS Department of Mathematics and Statistics.  

Abstract:

Local D-optimal experimental designs for precise parameter estimation minimize the determinant of the variance-covariance matrix of the parameter estimates based on a linear approximation of the model function. For nonlinear models, this determinant may not give a true indication of the volume of the parameter joint inference regions and therefore, the resulting designs could be non-informative. In this presentation, a new formulation of the D-optimality using profile-based sensitivity coefficients developed by Sulieman et.al. (2009) is discussed. The newly formulated D-optimal criterion accounts for both parameter co-dependencies and model nonlinearity and is shown to reduce parameter estimate correlations while improving their precision. Model examples are illustrated to show the applicability of the new bi-objective design criterion.

For further details, kindly contact [email protected].