mlegp (version 2.2.6)

createGP: creates a Gaussian process object

Description

creates a Gaussian process gp object

Usage

createGP(X, Z, beta, a, meanReg, sig2, nugget, param.names = 1:dim(X)[2], constantMean = 1)

Arguments

X

the design matrix

Z

output obtained from the design matrix X, as a vector or a 1-column matrix

beta

vector of correlation coefficients

a

vector of smoothness parameters in the correlation function (if mlegp is used, these will be 2)

meanReg

the constant mean if constantMean = 1, otherwise the regression coefficients of the mean function such that meanReg pre-multiplied by (1 X) will produce the mean matrix

sig2

the unconditional variance of the Gaussian process

nugget

the constant nugget or a vector of length nrow(X) corresponding to the diagonal nugget matrix

param.names

optional vector of parameter names (with length equal to ncol(X)

constantMean

1 if the Gaussian process has a constant mean; 0 otherwise

Value

• an object of class gp that contains the following components:
• Z
  matrix of observations
• numObs
  number of observations
• X
  the design matrix
• numDim
  number of dimensions of X
• constantMean
  1 if GP has a constant mean; 0 otherwise
• mu
  the mean matrix
• Bhat
  mean function regression coefficients
• beta
  correlation parameters
• a
  smoothness parameters in correlation function
• sig2
  unconditional variance of predicted expected output
• params
  vector of parameter names, corresponding to columns of X
• invVarMatrix
  inverse var-cov matrix
• nugget
  constant nugget or vector corresponding to the diagonal nugget matrix
• loglike
  the log likelihood of the observations
• cv
  results from cross-validation, where cv[,1] are the cross-validated predictions cv[,2] are the variances of the cross-validated predictions

References

http://users.nsula.edu/dancikg/mlegp/

See Also

mlegp