Skip to main content
. 2018 Dec 5;474(2220):20180400. doi: 10.1098/rspa.2018.0400

Figure 5.

Figure 5.

The difference of the cross-validated sample quantile and the corresponding standard Gaussian theoretical quantile (qsampleqtheory) plotted against the theoretical quantile (qtheory) for LOOO cross-validation at 300 dbar. The compared models are the RG-like reference model (model 1), the locally stationary one-month spatial and three-month spatio-temporal models with a Gaussian nugget (models 2 and 5) and the corresponding models with a Student nugget (models 3 and 6). The closer the curves are to a horizontal straight line at 0, the better the calibration of the predictive distributions.