INLAMRA: INLA-MRA model for inference and prediction in spatiotemporal...

In villandre/MRAinla: IS-MRA: An algorithm for large spatiotemporal datasets

Description

Fits the INLA-MRA model to a spatiotemporal dataset.

Usage

INLAMRA(
  responseVec,
  covariateFrame = NULL,
  spatialCoordMat,
  timePOSIXorNumericVec,
  predCovariateFrame = NULL,
  predSpatialCoordMat = NULL,
  predTimePOSIXorNumericVec = NULL,
  sinusoidalProjection = FALSE,
  spatialRangeList = NULL,
  spatialSmoothnessList = list(start = log(1.5)),
  timeRangeList = NULL,
  timeSmoothnessList = list(start = log(0.5)),
  scaleList = list(start = 0, hyperpars = c(mu = 0, sigma = 2)),
  errorSDlist = list(start = 0),
  fixedEffSDlist = list(start = log(10)),
  FEmuVec = rep(0, ncol(covariateFrame) + 1),
  control = NULL
)

Arguments

responseVec	A numeric vector with response values.
covariateFrame	A data.frame containing covariate values in the order of elements in responseVec. Character elements should be re-coded as factors.
spatialCoordMat	A matrix or data.frame with two columns with the first corresponding to longitude, and the second to latitude; can also be x and y if the sinusoidal projection is used.
timePOSIXorNumericVec	A vector of time values, preferrably in POSIX* format. Can also be in numeric format.
predCovariateFrame	A data.frame containing covariate values for the prediction datasets.
predSpatialCoordMat	Like spatialCoordMat, but for the prediction data.
predTimePOSIXorNumericVec	Like timePOSIXorNumericVec, but for prediction data.
sinusoidalProjection	Logical value indicating whether the provided coordinates are in the sinusoidal projection.
spatialRangeList	List with two elements: a starting value for the spatial range log-hyperparameter, and a two element vector giving the mean and standard deviation of the normal hyperprior (second element must be omitted if hyperparameter is fixed).
spatialSmoothnessList	List with two elements: a starting value for the spatial smoothness log-hyperparameter, and a length-two vector giving the mean and standard deviation of the associated normal hyperprior (second element must be omitted if hyperparameter is fixed.
timeRangeList	List with two elements: a starting value for the temporal range log-hyperparameter, and a length-two vector giving the mean and standard deviation of the associated normal hyperprior (second element must be omitted if hyperparameter is fixed).
timeSmoothnessList	List with two elements: a starting value for the temporal smoothness log-hyperparameter, and a length-two vector giving the mean and standard deviation of the associated normal hyperprior (second element must be omitted if hyperparameter is fixed).
scaleList	List with two elements: a starting value for the scale log-hyperparameter, and a length-two vector giving the mean and standard deviation of the associated normal hyperprior (second element must be omitted if hyperparameter is fixed)
errorSDlist	List with two elements: a starting value for the uncorrelated error standard deviation log-hyperparameter, and a length-two vector giving the mean and standard deviation of the associated normal hyperprior (second element must be omitted if hyperparameter is fixed).
fixedEffSDlist	List with two elements: a starting value for the uncorrelated fixed effects standard deviation log-hyperparameter, and a length-two vector giving the mean and standard deviation of the associated normal hyperprior (second element must be omitted if hyperparameter is fixed).
FEmuVec	Vector with the mean value of the priors for the fixed effects. Its length must be equal to the number of columns in covariateFrame plus one (for the intercept).
control	List of control parameters. See ?INLAMRA.control.

Details

Spatial coordinates must use the longitude/latitude ("+proj=longlat +datum=WGS84 +no_defs +ellps=WGS84 +towgs84=0,0,0") or sinusoidal ("+proj=sinu +lon_0=0 +x_0=0 +y_0=0 +R=6371007.181 +units=m +no_defs") projection. In the latter case, they will be automatically converted to the lon./lat. projection.

Some of the control parameters should be tuned to ensure better computational or predictive performance, or to make it possible to stop and resume model fitting. See INLAMRA.control.

Value

A list with three components:

• hyperMarginalMoments A data.frame giving the mean, and standard deviation of the marginal log-hyperparameter posteriors, as well as their 95% credible intervals. Note that the scaling of the time hyperparameters depends on the provided time values. If they are inputted as POSIX* objects, time hyperparameters will relate to time measured in days. If they are inputted as numeric, the original scale is used instead.

• FEmarginalMoments A data.frame giving the mean and standard deviation of the marginal fixed effects posteriors, as well as their 95% credibility intervals.

• predMoments A data.frame with two columns, Mean and SD. The order of the predictions matches the one in predCovariateFrame.

Examples

## Not run: 
# See example in vignette.

## End(Not run)

villandre/MRAinla documentation built on April 20, 2021, 7:49 a.m.