docopulae (version 0.2.1)

FedorovWynn: Fedorov Wynn

Description

FedorovWynn finds a D- or Ds-optimal design using a Fedorov-Wynn-type algorithm.

Usage

FedorovWynn(mod, sNames = NULL, names = NULL, tolAbs = Inf,
  tolRel = 1e-04, maxIter = 10000)

Arguments

mod

some model.

sNames

a vector of names or indices, the subset of parameters to optimize for.

names

a vector of names or indices, the set of parameters use.

tolAbs

the absolute tolerance regarding the sensitivities.

tolRel

the relative tolerance regarding the sensitivities with respect to the theoretical limit.

maxIter

the maximum number of iterations.

Value

• FedorovWynn returns an object of class "desigh". An object of class "desigh" is a list containing the following components:
  • mod: argument
  • x: a row matrix of design points, here identical tomod$x.
  • w: a numeric vector of weights, for each design point respectively.
  • sens: a numeric vector of sensitivities, for each design point respectively.
  • args: a list of arguments.
  • adds: a list of additional (runtime) information.

Details

Both sNames and names default to the set of parameters for which the Fisher information is available.

The algorithm starts from a uniform weight design. In each iteration weight is redistributed to the point which has the highest sensitivity. Sequence: 1/i. The algorithm stops when all sensitivities are below a certain absolute and relative tolerance level, or the maximum number of iterations is reached.

References

Fedorov, V. V. (1971) The Design of Experiments in the Multiresponse Case. Theory of Probability and its Applications, 16(2):323-332.

Wynn, Henry P. (1970) The Sequential Generation of D-Optimum Experimental Designs. The Annals of Mathematical Statistics, 41(5):1655-1664.

See Also

param, getM, reduce, plot.desigh, Defficiency, update.desigh

Examples

## see examples for param