SAVE-class: SAVE class
In SAVE: Bayesian Emulation, Calibration and Validation of Computer Models

Description Objects from the Class Slots Methods Author(s) See Also

S4 class for Statistical Analysis and Validation Engine.

Objects of this class are created and initialized with the function SAVE that computes the calculations needed for setting up the analysis. These can be completed with the function bayesfit that performs the Bayesian analysis in the SAVE methodology

responsename:

Object of class character. The response name.

controllablenames:

Object of class character. The names of the controllable inputs.

calibrationnames:

Object of class character. The names of the calibration inputs.

constant.controllables:

Object of class logical. Controls whether or not the analysis has constant controllable inputs.

df:

Object of class matrix. The field design once the replicates (if any) have been removed.

dm:

Object of class matrix. The model design.

ym:

Object of class numeric. Model response associated with dm.

yf:

Object of class numeric. The field observations.

meanformula:

Object of class formula. The formula that specifies the mean function of the emulator of the computer model.

mle:

The maximum likelihood estimates. This is a list with three components

thetaM:: A numeric vector containing the estimate of the parameters specifying the covariance structure of the emulator of the computer model. This covariance function has precision lambdaM and a separable correlation function with k(x,y)=exp(-betaM*h^alphaM) where h=abs(x-y). The vector thetaM is organized as follows: (lambdaM, betaM, alphaM), where betaM and alphaM are named vectors.
thetaL:: The numeric vector of regression coefficients associated with the mean function of the emulator of the computer model
thetaF:: A numeric vector organized as (lambdaB, betaB, alphaB, lambdaF) containing the estimates of lambdaF, the precision of the field measurement error, and of the parameters specifying the Gaussian process prior of the bias function. The covariance function and the parameters follow the same structure as that described for thetaM

bestguess:

A numeric vector containing the best guess (provided in the call) for the calibration inputs.

xm:

The model matrix corresponding to the evaluation of the meanformula at dm.

xf:

The model matrix corresponding to the evaluation of the meanformula at df.

prior:

Description of the prior used (empty if if bayesfit is not run).

method:

A numeric object with possible values 1 and 2. Two different MCMC methods have been implemented. If method=2 then the computer model and bias are integrated out (analytically) before sampling the calibration parameters. If method=1 then the calibration parameters is sampled from the full conditional. (Empty if if bayesfit is not run).

mcmcMultmle:

A positive numeric object. Priors for the precisions (lambdaM and lambdaB) are exponential distributions centered at the corresponding mle multiplied by mcmcMultmle. (Empty if if bayesfit is not run).

mcmcsample: