|Friday, February 16|
|CS02 Practical Considerations for Modeling||
Fri, Feb 16, 9:15 AM - 10:45 AM
Flexible Modeling and Experimental Design Strategies (303520)*Timothy E. O'Brien, Loyola University Chicago
Keywords: design, dose-response, growth curves, logistic regression, statistical modelling
Researchers often find that nonlinear regression models are more applicable for modelling various physical processes than are linear ones since they tend to fit the data well and since these models are more scientifically meaningful. For example, researchers in fields as diverse as toxicology and pharmacology, biometry, economics, education and psychology typically fit four-parameter sigmoidal functions, and are often in a position of requiring optimal or near-optimal designs for the chosen nonlinear model. A common shortcoming of most optimal designs for nonlinear models used in practical settings, however, is that these designs typically focus only on (first-order) parameter variance or predicted variance, and thus ignore the inherent nonlinear of the assumed model function. Another shortcoming of optimal designs is that they typically cannot be used to test for model adequacy.
This talk reviews and underscores the practical advantages of nonlinear models, and examines various robust design criteria, including geometric and uniform design strategies and those based on second-order considerations. Key examples are provided to illustrate these ideas, using SAS and R software.