createSTmodel: Construct STmodel Object
In SpatioTemporal: Spatio-Temporal Model Estimation

Description Usage Arguments Details Value Author(s) See Also Examples

Creates a STmodel object that can be for estimation and prediction. For details see the sub-functions linked under the relevant Arguments.

createSTmodel(STdata, LUR = NULL, ST = NULL, cov.beta = list(covf = "exp",
  nugget = FALSE), cov.nu = list(covf = "exp", nugget = TRUE, random.effect =
  FALSE), locations = list(coords = c("x", "y"), long.lat = NULL, coords.beta
  = NULL, coords.nu = NULL, others = NULL), strip = FALSE, scale = FALSE,
  scale.covars = NULL)

`STdata`	`STdata` object with observations, covariates, trends, etc; see `createSTdata` or `mesa.data.raw` for an example.
`LUR`	Specification of covariates for the beta-fields, see `processLUR`.
`ST`	Specification of spatio-temporal covariates, see `processST`.
`cov.beta, cov.nu`	Specification of the covariance functions, see `updateCovf`.
`locations`	Specification of the sites (both monitored and un-monitored), see `processLocation`.
`strip`	Should unobserved locations be dropped?
`scale`	Scale the covariates? If `TRUE` all non-factor covariates are scaled after the locations have been extracted but before constructing the covariate matrix for the beta-fields. (NOTE: If set to `TRUE` this scales the `LUR.all` elements to mean=0, sd=1).
`scale.covars`	list with elements `mean` and `sd` giving the mean and standard deviation to use when scaling the covariates. Computed from `STdata$covars` if not given.

The object holds observations, trends, geographic, and spatio-temporal covariates, as well as a number of precomputed fields that speed up log-likelihood evaluations. To improve performance the locations are also reorder so that observed locations come before unobserved.

A STmodel object, see mesa.model for an example.

Johan Lindstrom

Other STmodel methods: MCMC.STmodel, c.STmodel, estimate.STmodel, estimateCV.STmodel, plot.STdata, predict.STmodel, print.STmodel, print.summary.STmodel, qqnorm.predCVSTmodel, scatterPlot.predCVSTmodel, simulate.STmodel, summary.STmodel

Other STmodel functions: createCV, createDataMatrix, dropObservations, estimateBetaFields, loglikeSTdim, loglikeST, predictNaive, processLUR, processLocation, updateCovf, updateTrend.STdata

Other STdata functions: c.STmodel, createDataMatrix, createSTdata, detrendSTdata, estimateBetaFields, removeSTcovarMean, updateTrend.STdata

##load the data
data(mesa.data.raw)
##and create STdata-object
mesa.data <- createSTdata(mesa.data.raw$obs, mesa.data.raw$X, n.basis=2,
                          SpatioTemporal=mesa.data.raw["lax.conc.1500"])

##define land-use covariates
LUR <-  list(~log10.m.to.a1+s2000.pop.div.10000+km.to.coast,
             ~km.to.coast, ~km.to.coast)
##and covariance model
cov.beta <- list(covf="exp", nugget=FALSE)
cov.nu <- list(covf="exp", nugget=TRUE, random.effect=FALSE)
##which locations to use
locations <- list(coords=c("x","y"), long.lat=c("long","lat"), others="type")

##create object
mesa.model <- createSTmodel(mesa.data, LUR=LUR, ST="lax.conc.1500",
                            cov.beta=cov.beta, cov.nu=cov.nu,
                            locations=locations)
print(mesa.model)
##This is the same as data(mesa.model)

##lets try some alternatives:
model.none <- createSTmodel(mesa.data, LUR=NULL, ST=NULL)
print(model.none)

##Specify LUR:s using numbers
names(mesa.data$covars)
model.diff <- createSTmodel(mesa.data, LUR=list(c(7,10,11,12),11:12,11:12),
                            ST=1)
print(model.diff)

##Same covariates for all temporal trends, calling by name
##but with different covariance models for each trend, and nugget that depends
##on monitor type
model.same <- createSTmodel(mesa.data, LUR=c("log10.m.to.a1", "log10.m.to.road",
                                         "km.to.coast","s2000.pop.div.10000"),
                            ST="lax.conc.1500", cov.nu=list(nugget="type"),
                            cov.beta=list(covf=c("exp","exp2","iid"),
                              nugget=c(FALSE, FALSE, TRUE)) )
print(model.same)

Loading required package: Matrix
STmodel-object with:
	No. locations: 25 (observed: 25)
	No. time points: 280 (observed: 280)
	No. obs: 4577

Trend with 2 basis function(s):
[1] "V1" "V2"
with dates:
	1999-01-13 to 2009-09-23

Models for the beta-fields are:
$const
~log10.m.to.a1 + s2000.pop.div.10000 + km.to.coast

$V1
~km.to.coast

$V2
~km.to.coast

1 spatio-temporal covariate(s):
[1] "lax.conc.1500"

Covariance model for the beta-field(s):
	Covariance type(s): exp, exp, exp 
	Nugget: No, No, No 
Covariance model for the nu-field(s):
	Covariance type: exp 
	Nugget: ~1 
	Random effect: No 
All sites:
  AQS FIXED 
   20     5 
Observed:
  AQS FIXED 
   20     5 

For AQS:
  Number of obs: 4178
  Dates: 1999-01-13 to 2009-09-23
For FIXED:
  Number of obs: 399
  Dates: 2005-12-07 to 2009-07-01
STmodel-object with:
	No. locations: 25 (observed: 25)
	No. time points: 280 (observed: 280)
	No. obs: 4577

Trend with 2 basis function(s):
[1] "V1" "V2"
with dates:
	1999-01-13 to 2009-09-23

Models for the beta-fields are:
$const
~1

$V1
~1

$V2
~1

No spatio-temporal covariates.

Covariance model for the beta-field(s):
	Covariance type(s): exp, exp, exp 
	Nugget: No, No, No 
Covariance model for the nu-field(s):
	Covariance type: exp 
	Nugget: ~1 
	Random effect: No 
 [1] "ID"                  "x"                   "y"                  
 [4] "long"                "lat"                 "type"               
 [7] "log10.m.to.a1"       "log10.m.to.a2"       "log10.m.to.a3"      
[10] "log10.m.to.road"     "km.to.coast"         "s2000.pop.div.10000"
STmodel-object with:
	No. locations: 25 (observed: 25)
	No. time points: 280 (observed: 280)
	No. obs: 4577

Trend with 2 basis function(s):
[1] "V1" "V2"
with dates:
	1999-01-13 to 2009-09-23

Models for the beta-fields are:
$const
~log10.m.to.a1 + log10.m.to.road + km.to.coast + s2000.pop.div.10000

$V1
~km.to.coast + s2000.pop.div.10000

$V2
~km.to.coast + s2000.pop.div.10000

1 spatio-temporal covariate(s):
[1] "lax.conc.1500"

Covariance model for the beta-field(s):
	Covariance type(s): exp, exp, exp 
	Nugget: No, No, No 
Covariance model for the nu-field(s):
	Covariance type: exp 
	Nugget: ~1 
	Random effect: No 
STmodel-object with:
	No. locations: 25 (observed: 25)
	No. time points: 280 (observed: 280)
	No. obs: 4577

Trend with 2 basis function(s):
[1] "V1" "V2"
with dates:
	1999-01-13 to 2009-09-23

Models for the beta-fields are:
$const
~log10.m.to.a1 + log10.m.to.road + km.to.coast + s2000.pop.div.10000

$V1
~log10.m.to.a1 + log10.m.to.road + km.to.coast + s2000.pop.div.10000

$V2
~log10.m.to.a1 + log10.m.to.road + km.to.coast + s2000.pop.div.10000

1 spatio-temporal covariate(s):
[1] "lax.conc.1500"

Covariance model for the beta-field(s):
	Covariance type(s): exp, exp2, iid 
	Nugget: No, No, Yes 
Covariance model for the nu-field(s):
	Covariance type: exp 
	Nugget: ~type 
	Random effect: No 
All sites:
  AQS FIXED 
   20     5 
Observed:
  AQS FIXED 
   20     5 

For AQS:
  Number of obs: 4178
  Dates: 1999-01-13 to 2009-09-23
For FIXED:
  Number of obs: 399
  Dates: 2005-12-07 to 2009-07-01