Reverting to using the likelihood sampling method. This has two benefits:
The Bayesian specification that D > 0 can be introduced
In future, using a non-linear relationship will be possible if desired
Having fixed the problems with the covariance matrix sensitivity previously, this should give the same result as the random samples from a Gaussian process method.
Reverting to using the likelihood sampling method. This has two benefits:
Having fixed the problems with the covariance matrix sensitivity previously, this should give the same result as the random samples from a Gaussian process method.