View source: R/Federov.design.R
Federov.paramset | R Documentation |
This function provides a wrapper for constructing Federov designs for parameters by using the parameter set logic of quantstrat and the experiment design functions from the 'AlgDesign' package. It will construct a full or approximate Federov design via 'method="Federov"' (not case sensitive) using the 'optFederov' function and will construct a constrained Monte Carlo Federov design via 'method="MonteCarlo"' using the 'optMonteCarlo' function.
Federov.paramset(
strategy.st,
paramset.label,
...,
method = "Federov",
printd = FALSE,
returnlist = FALSE,
approximate = TRUE,
center = TRUE,
constrain = FALSE
)
strategy.st |
astring describing the name of an object of type 'strategy' that contains a parameter set to construct a Federov design for |
paramset.label |
label describing the paramset to use in the strategy object |
... |
any other passthrough parameters |
method |
one of "Federov" or "MonteCarlo", see Details |
printd |
if TRUE, print the design summary |
returnlist |
if TRUE, return the list object describing the Federov design, else return just the 'param.combos' |
approximate |
if FALSE, use an exact design, will be slower but more accurate than if TRUE |
center |
if TRUE, the default, center the parameters around a center value, see Details |
constrain |
if TRUE, 'apply.constraints' for the full Federov design, see Details |
It is important to note from the beginning that while 'method="Federov"' is the default, that the closed form Federov design is not suitable when there are constraints. We will warn the user if there are constraints in the strategy specification and a full Federov design is chosen. The function also supports the option 'constrain=TRUE' that will apply constraints to the full Federov design. For some strategies with many constraints, this may result in a significantly unbalanced set. This should still be OK as a starting point for optimization, but may hamper some statistical inference about parameter interactions and lower the overall power of the design if the unbalanced nature is severe.
For now, for constrained Monte Carlo Federov designs, we are not supporting factor or mixture models, though these designs are supported by the 'AlgDesign' package. Patches welcome, or even just discussion of solid use cases.
The user may use ... to pass through any additional parameters for use by 'optFederov' or 'optMonteCarlo'. For example, 'nLevels', 'nCand', 'nRepeats' may be candidates for finer grained control. see '?AlgDesign::optMonteCarlo' for details. We have tried to make reasonable decisions based on the data contained in the strategy object. 'nLevels' defaults to the minimum of 5 (though 3 is a reasonable and smaller choice) or the number of levels contained in the paramset for that variable.
optFederov
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.