staircase.hyper.est:
Estimate hyperparameters of ungauged sites
Description
This function combines the results from the staircase.EM fit and the
SG method to estimate the hyperparameters associated with the ungauged sites.
Usage
staircase.hyper.est(emfit, covfit, u, p, g, d0 = NULL)
Arguments
emfit
Output from the staircase.EM fit
covfit
The covariance matrix between all locations (with new locations
at the beginning). This is an output from the SG fitting
u
number of new locations
p
dimension of the multivariate response
d0
(optional) The degrees of freedom for the new locations (ungauged block)
Value
List with the following elements:
- Delta.0
-
The degree of freedoms for the new locations. Equal to
d0, if given (must be $> u*p+2$),
else
mean(emfit$delta) if $> u*p+2$;
u*p+ min(emfit$delta) otherwise.
- Lambda.0
-
Conditional variance between new locations given the gauged
stations
- Xi0.0
-
the regression slope (Note: $\tau_{0i} = {\rm kronecker}(\xi_0 ,
diag(p))$)
- H.0
-
The variance matrix for the rows of $\tau^{[u]}$
Also all components of the output of the staircase.EM fit (for
blocks 1-K).