Changelog
Source:NEWS.md
rSPDE (development version)
- Created the
group_predict
function, to obtain predictions on a testing set based on observations on a training set. - Added support for
stochvol
,stochvol.nig
,stochvolln
andbinomial
likelihoods incross_validation
function. - Changing the default
nu.upper.bound
to 2 in dimension 1, and keeping the defaultnu.upper.bound
to 4 in dimension 2 inrspde.matern()
function. - Created
matern.rational()
operators for creating stationary matern operators. - Created
spacetime.operators()
for creating space-time models. - Created
matern2d.operators()
for anisotropic operators. - Implemented space-time operators in cgeneric to be used in
INLA
andinlabru
. - Added vignette on space-time models.
- Added vignette on stationary models.
- Added vignette on anisotropic models.
rSPDE 2.3.3
CRAN release: 2023-11-05
- Bugfix on rspde_lme when fitting with fixed smoothness.
- Added a 2d fem interface.
- Moved from using INLA’s mesh functions to fmesher’s mesh functions.
- Removing rgdal from suggests.
- The
data
argument inpredict.rspde_lme
has been changed tonewdata
. - Adding
covariance_mesh
andcov_function_mesh
methods as functions in the list returned by objects obtained frommatern.operators()
andspde.matern.operators()
. - Updated the internal structure to match the updates from the
MetricGraph
package. - Updated the
cross_validation
function to match the updates ininlabru
. - Added
glance
andaugment
methods forrspde_lme
objects.
rSPDE 2.3.2
CRAN release: 2023-07-02
- Small improvement on speed for rspde_lme.
- Bugfix on Q for small values of nu in dimension 1.
- Adding parameterization option for rspde.result.
- Bugfix on which_repl in rspde_lme.
- Addressing issues related to the new version of the Matrix package.
rSPDE 2.3.1
CRAN release: 2023-05-25
- Adding references in DESCRIPTION.
- Changing link to eigen library.
rSPDE 2.3.0
- Fixed a bug on rSPDE.construct.matern.loglike when the parameterization is “matern”.
- Created the rspde_lme() interface, with corresponding standard methods(predict, summary, etc).
- Updated the vignettes to use the rspde_lme() interface instead of the likelihood function factory.
- Replaced chol by Cholesky when using it to compute determinants or to solve systems.
rSPDE 2.2.0
CRAN release: 2023-04-12
- Adding a new parameterization (variance and a range-like parameter)
- Posterior sampling on the predict method.
- Added the
cross_validation
function which has several scoring rules implemented (MSE, CRPS, SCRPS, DSS) based on ourinlabru
implementation of the rational SPDE approach.
rSPDE 2.1.0
CRAN release: 2023-01-19
- Expanded the parameterization options on matern.operators and spde.matern.operators, along with their associated functions.
- Implementation of the precision method for inla_rspde objects.
- Implementation of the covariance-based spde.matern.operators function and its associated functions.
- Adjusts on the compatibility with the forthcoming MetricGraph package.
rSPDE 2.0.0
- Added cgeneric versions of the nonstationary models
- Added support for metric graphs (depends on the MetricGraph package)
- Added cgeneric versions of the stationary models
- Replaced rgeneric models by their cgeneric counterparts
- Added a new parameterization (range and std. dev)
- Created a new method gg_df to help posterior plotting in ggplot2
rSPDE 1.2.0
CRAN release: 2022-09-16
- Added an inlabru interface
- Added “rational.order” and “rational.type” functions
- Added the BRASIL rational approximation
- Improved covariance-based operator objects
- Improved log-likelihood computation
- Created 2d folded Matern under different boundary conditions
- Implemented different boundary conditions for 1d folded Matern
rSPDE 1.1.0
- Minor typos on vignettes and man pages were corrected
- Some examples were changed to improve their numerical stability
rSPDE 1.0.0
CRAN release: 2021-12-13
- Implementation of the covariance-based rational approximation for stationary Matérn models
- R-INLA implementation of the rational SPDE approach
- Added an introduction to rSPDE vignette
- The previous vignette was updated an became an operator-based rational approximation vignette
- Added a vignette for the R-INLA implementation of the SPDE approach
- Added a vignette to present the rational approximation using the rSPDE package
- Backward compatibility was maintained