mvcokm-class {ARCokrig} | R Documentation |

This is an S4 class definition for `mvcokm`

in the `ARCokrig`

package

`output`

a list of

*s*elements, each of which contains a matrix of computer model outputs.`input`

a list of

*s*elements, each of which contains a matrix of inputs.`param`

a list of

*s*elements, each of which contains a vector of initial values for correlation parameters (and nugget variance parameters if nugget terms are included in AR-cokriging models).`cov.model`

a string indicating the type of covariance function in AR-cokriging models. Current covariance functions include

- exp
product form of exponential covariance functions.

- matern_3_2
product form of Matern covariance functions with smoothness parameter 3/2.

- matern_5_2
product form of Matern covariance functions with smoothness parameter 5/2.

- Gaussian
product form of Gaussian covariance functions.

- powexp
product form of power-exponential covariance functions with roughness parameter fixed at 1.9.

- aniso_exp
anisotropic form of exponential covariance function.

- aniso_matern_3_2
anisotropic form of Matern covariance functions with smoothness parameter 3/2.

- aniso_matern_5_2
anisotropic form of Matern covariance functions with smoothness parameter 5/2.

`nugget.est`

a logical value indicating whether the nugget is included or not. Default value is

`FALSE`

.`prior`

a list of arguments to setup the prior distributions with the jointly robust prior as default

- name
the name of the prior. Current implementation includes

`JR`

,`Reference`

,`Jeffreys`

,`Ind_Jeffreys`

- hyperparam
hyperparameters in the priors. For jointly robust (JR) prior, three parameters are included:

*a*refers to the polynomial penalty to avoid singular correlation matrix with a default value 0.2;*b*refers to the exponenetial penalty to avoid diagonal correlation matrix with a default value 1; nugget.UB is the upper bound of the nugget variance with default value 1, which indicates that the nugget variance has support (0, 1).

`opt`

a list of arguments to setup the

`optim`

routine.`NestDesign`

a logical value indicating whether the experimental design is hierarchically nested within each level of the code.

`tuning`

a list of arguments to control the MCEM algorithm for non-nested design. It includes the arguments

- maxit
the maximum number of MCEM iterations.

- tol
a tolerance to stop the MCEM algorithm. If the parameter difference between any two consecutive MCEM algorithm is less than this tolerance, the MCEM algorithm is stopped.

- n.sample
the number of Monte Carlo samples in the MCEM algorithm.

`info`

a list that contains

- iter
number of iterations used in the MCEM algorithm

- eps
parameter difference after the MCEM algorithm stops

Pulong Ma <mpulong@gmail.com>

[Package *ARCokrig* version 0.1.1 Index]