We consider the problem of constructing D-optimal designs with covariates that are fixed and specified in advance. Specifically, we present methods for computing (truncated) optimal designs when nuisance parameters exist and interest focuses on a subset of the parameters in a linear model. Relationships between truncated optimal designs and optimal designs when nuisance parameters do not exist are characterized.
- Additive models
- constrained optimal design
- marginally restricted designs
- nuisance parameters
- truncated optimal design