plot.minimax {ICAOD}

Plotting minimax Objects

Description

This function plots the evolution of the ICA algorithm (iteration vs the best (minimum) criterion value at each iteration) and also verifies the optimality of the last obtained design using the general equivalence theorem. It plots the sensitivity function and calculates the ELB for the best design generated at iteration number iter.

Usage

## S3 method for class 'minimax'
plot(
  x,
  iter = NULL,
  sensitivity = TRUE,
  calculate_criterion = FALSE,
  sens.minimax.control = list(),
  crt.minimax.control = list(),
  sens.bayes.control = list(),
  crt.bayes.control = list(),
  sens.control = list(),
  silent = TRUE,
  plot_3d = c("lattice", "rgl"),
  evolution = FALSE,
  ...
)

Arguments

x

An object of class minimax.

iter

Iteration number. if NULL (default), it will be set to the last iteration.

sensitivity

Logical. If TRUE (default), the general equivalence theorem is used to check the optimality if the best design in iteration number iter and the sensitivity function will be plotted.

calculate_criterion

Logical. Re-calculate the criterion value (maybe with a set of new tuning parameters to be sure of the globality of the maximum over the parameter space given the design)? It only assumes a continuous parameter space for the minimax and standardized maximin designs. Defaults to FALSE. See 'Details'.

sens.minimax.control

Control parameters to verify general equivalence theorem. For details, see sens.minimax.control. If NULL (default), it will be set to the tuning parameters used to create object x.

crt.minimax.control

Control parameters to optimize the minimax or standardized maximin criterion at a given design over a continuous parameter space. For details, see crt.minimax.control.

sens.bayes.control

Control parameters to verify general equivalence theorem for the Bayesian optimal designs. For details, see sens.bayes.control. If NULL (default), it will be set to the tuning parameters used to create object x.

crt.bayes.control

Control parameters to optimize the integration in the Bayesian criterion at a given design over a continuous parameter space. For details, see crt.bayes.control. If NULL (default), it will be set to the tuning parameters used to create object x. If NULL (default), it will be set to the tuning parameters used to create object x.

sens.control

Control Parameters for Calculating the ELB. For details, see the function sens.control.

silent

Do not print anything? Defaults to TRUE.

plot_3d

Which package should be used to plot the sensitivity function for two-dimensional design space. Defaults to plot_3d = "lattice". Only applicable when sensitivity = TRUE.

evolution

Plot Evolution? Defaults to FALSE.

...

Argument with no further use.

Details

In addition to verifying the general equivalence theorem, this function makes it possible to re-calculated the criterion value for the output designs using a new set of tuning parameters, especially, a large value for maxeval in the function crt.minimax.control. This is useful for minimax and standardized maximin optimal designs to assess the robustness of the criterion value with respect to different values of maxeval. To put it simple, for these designs, the user can re-calculate the criterion value (finds the global maximum over the parameter space given an output design in a minimax problem) with larger values for maxeval in crt.minimax.control to be sure that the function nloptr finds global optima of the inner optimization problem over the parameter space using the default value (or the user-given value) of maxeval. If increasing the value of maxeval returns different criterion values, then the results can not be trusted and the algorithm should be repeated with a higher value for maxeval.

See Also

minimax, locally, robust