makePrediction: Spatial projected sequential GP prediction

Description Usage Arguments Details Warning Author(s) References See Also Examples

View source: R/makePrediction.R


makePrediction performs prediction/interpolation within the PSGP package.


 makePrediction(object, vario) 



a list object of intamap type. Most arguments necessary for interpolation are passed through this object. See intamap-package for further description of the necessary content of this variable. Additional meta data about the measurement process is included in this object. In particular, see learnParameters for a way to specify measurement error variances.


Log-parameters of the covariance function. For compatibility with the intamap package, the log-parameters of the PSGP covariance function are stored within a variogram array object (see vgm), as follows:
vario[1,1] NA
vario[1,2] length scale (or range) of the Exponential kernel
vario[1,3] variance (or sill) of the Exponential kernel
vario[1,4] length scale (or range) of the Matern 5/2 kernel
vario[1,5] variance (or sill) of the Matern 5/2 kernel
vario[1,6] inverse bias (i.e. 1/mean(data))
vario[1,7] white noise variance (nugget)


The Projected Spatial Gaussian Process (PSGP) framework provides an approximation to the full Gaussian process in which the observations are projected sequentially onto an optimal subset of 'active' observations. Spatial interpolation is done using a mixture of covariance kernels (Exponential and Matern 5/2).

The function makePrediction is a function for making predictions at a set of unobserved inputs (or locations).

Measurement characteristics (i.e. observation error) can be specified if needed. See learnParameters for a description of how to specify measurement error models with given variances.


It is advised to use the intamap wrapper spatialPredict rather than calling this method directly.


Ben Ingram, Remi Barillec


L. Csato and M. Opper. Sparse online Gaussian processes. Neural Computation, 14(3): 641-669, 2002.

B. Ingram, D. Cornford, and D. Evans. Fast algorithms for automatic mapping with space- limited covariance functions. Stochastic Environmental Research and Risk Assessment, 22 (5):661-670, 2008.

See Also

learnParameters spatialPredict, createIntamapObject


  # see example in spatialPredict

psgp documentation built on Feb. 1, 2020, 1:07 a.m.