Creates a hyperparameter object from a vector of length 19. Intended as a toy example to be modified for real-world cases.
hpa.fun.toy(x)
The four sigmas corresponding to the four levels
The four B matrices corresponding to the four levels
The three (sic) matrices corresponding to levels 1-3
Vector of length 19 that specifies the correlation scales
Robin K. S. Hankin
Elements 1-4 of x specify the sigmas for each of the four levels in the toy example. Elements 5-7 specify the correlation scales for level 1, elements 8-10 the scales for level 2, and so on.
Internal function pdm.maker()
shows how the B matrix is
obtained from the various elements of input argument x
. Note
how, in this simple example, the B matrices are diagonal, but
generalizing to non-diagonal matrices should be straightforward (if
you can guarantee that they remain positive definite).
M. C. Kennedy and A. O'Hagan 2000. “Predicting the output from a complex computer code when fast approximations are available” Biometrika, 87(1): pp1-13