| Title: | AGricultural PRoductivity in Space |
| Version: | 2.0 |
| Description: | Functionalities to simulate space-time data and to estimate dynamic-spatial panel data models. Estimators implemented are the BCML (Elhorst (2010), <doi:10.1016/j.regsciurbeco.2010.03.003>), the MML (Elhorst (2010) <doi:10.1016/j.regsciurbeco.2010.03.003>) and the INLA Bayesian estimator (Lindgren and Rue, (2015) <doi:10.18637/jss.v063.i19>; Bivand, Gomez-Rubio and Rue, (2015) <doi:10.18637/jss.v063.i20>) adapted to panel data. The package contains functions to replicate the analyses of the scientific article entitled "Agricultural Productivity in Space" (Baldoni and Esposti (2021), <doi:10.1111/ajae.12155>)). |
| License: | GPL-3 |
| Encoding: | UTF-8 |
| LazyData: | true |
| RoxygenNote: | 7.1.1 |
| Depends: | R (≥ 2.10) |
| Imports: | methods, Matrix, plyr, sp, spdep, spacetime, matrixcalc, maxLik |
| Suggests: | INLA, terra |
| NeedsCompilation: | no |
| Packaged: | 2023-06-08 06:36:56 UTC; edoar |
| Author: | Edoardo Baldoni 
[aut, cre] |
| Maintainer: | Edoardo Baldoni <edoardo.baldoni@gmail.com> |
| Repository: | CRAN |
| Date/Publication: | 2023-06-08 07:32:54 UTC |
BCML estimator
Description
This function estimates a space time linear model according to the specified formula. It implements the BCML (or BCLSDV) estimator as in Elhorst (2010) doi:10.1016/j.regsciurbeco.2010.03.003.
Usage
bcml(
dataset,
yearStart,
yearEnd,
var.agg = "Cod_Provincia",
eq,
spatial = NULL,
estimation = "analytical",
corrBIAS = TRUE,
WMAT = NULL
)
Arguments
dataset |
STFDF with the data |
yearStart |
First year considered in the estimation |
yearEnd |
Last Anno considered in the estimation |
var.agg |
Index of the spatial units |
eq |
Formula to be estimated. It excludes the spatial lag |
spatial |
Radius to define neighbors |
estimation |
Either 'analytical' or 'numerical'. If 'analytical' is specified the concentrated maximum likelihood is estimated and then the other parameters are obtained analytically. Otherwise, all parameters are obtained through numerical maximization of the log-likelihood function. |
corrBIAS |
Boolean. If TRUE, the bias correction is applied. |
WMAT |
The spatial weight matrix |
Value
A list with two objects. The first object is the estimates table. The second object is the log-likelihood evaluated at its maximum
Examples
library(maxLik)
library(matrixcalc)
set.seed(123)
sd = sim_data_fe(dataset=regsamp,N=100,TT=8,
spatial = 80,Tau = -0.2,Rho = 0.4,
Beta = 2,sdDev = 2,startingT = 10,
LONGLAT = TRUE)
est_bcml = bcml(dataset = sd[[1]],yearStart = 3,yearEnd = 9,
var.agg = 'Cod_Provincia',eq = Y~X1,
estimation = 'analytical',corrBIAS = TRUE,WMAT = sd[[2]])
est_bcml
Matrix of technological distance
Description
A matrix of agricultural technological distance of NUTS3. It is used to weight the geographical distance of NUTS3 regions.
Usage
distOte
Format
A 106x106 matrix
Space-time bayesian INLA estimator
Description
This function estimates a space time linear model using the bayesian INLA. It is a wrapper of the INLA::inla function (Lindgren and Rue (2015) doi:10.18637/jss.v063.i19; Bivand, Gomez-Rubio and Rue (2015) doi:10.18637/jss.v063.i20) adapted to panel data.
Usage
inla.st(
formula,
d,
W,
RHO,
PHI,
var.agg,
normalization = FALSE,
improve = TRUE,
fhyper = NULL,
probit = FALSE,
...
)
Arguments
formula |
Formula of the model to be estimated |
d |
Data frame |
W |
Spatial matrix |
RHO |
Parameter of spatial dependence |
PHI |
Parameter of temporal dependence |
var.agg |
Indexes of the panel dimensions. The first argument is the spatial dimension, the second argument is the temporal dimension. |
normalization |
Boolean. If TRUE the data are normalized before estimation |
improve |
Please refer to the documentation of the INLA package |
fhyper |
Plase refer to the documentation of the INLA package |
probit |
Plase refer to the documentation of the INLA package |
... |
additional parameters. Please, refer to the documentation of the INLA package |
Value
Returns a model of class "inla". Please, refer to the documentation of the INLA package for additional information
Examples
library(terra)
set.seed(123)
sd = sim_data_fe(dataset=regsamp,N=100,TT=8,spatial = 80,
Tau = -0.2,Rho = 0.4, Beta = 2,sdDev = 2,
startingT = 10,LONGLAT = TRUE)
est_inla = inla.st(formula = Y~-1+X1,d = sd[[1]]@data,
W = sd[[2]],PHI=-0.2,RHO=0.4,
var.agg=c('Cod_Provincia','Anno'),
family='gaussian',
improve=TRUE,
normalization=FALSE,
control.family = list(hyper = list(prec=list(initial=25,fixed=TRUE))),
control.predictor = list(compute = TRUE),
control.compute = list(dic = TRUE, cpo = TRUE),
control.inla = list(print.joint.hyper = TRUE))
summary(est_inla)
Cropland grid of Northen Italy (20km x 20km squares)
Description
SpatialPolygonsDataFrame object with croplands of Northern Italy approximated with 20km x 20km squared polygons. It is based on croplands data contained in the Corine Land Cover 2012 raster map.
Usage
map1
Format
SpatialPolygonsDataFrame object
Source
https://land.copernicus.eu/pan-european/corine-land-cover/clc-2012
Cropland grid of Northen Italy (40km x 40km squares)
Description
SpatialPolygonsDataFrame object with croplands of Northern Italy approximated with 40km x 40km squared polygons. It is based on croplands data contained in the Corine Land Cover 2012 raster map.
Usage
map2
Format
SpatialPolygonsDataFrame object
Source
https://land.copernicus.eu/pan-european/corine-land-cover/clc-2012
Cropland grid of Northen Italy (60km x 60km squares)
Description
SpatialPolygonsDataFrame object with croplands of Northern Italy approximated with 60km x 60km squared polygons. It is based on croplands data contained in the Corine Land Cover 2012 raster map.
Usage
map3
Format
SpatialPolygonsDataFrame object
Source
https://land.copernicus.eu/pan-european/corine-land-cover/clc-2012
Cropland grid of Northen Italy (100km x 100km squares)
Description
SpatialPolygonsDataFrame object with croplands of Northern Italy approximated with 100km x 100km squared polygons. It is based on croplands data contained in the Corine Land Cover 2012 raster map.
Usage
map4
Format
SpatialPolygonsDataFrame object
Source
https://land.copernicus.eu/pan-european/corine-land-cover/clc-2012
MML estimator
Description
This function estimates a space time linear model according to the specified formula using the ML estimator as in Elhorst (2010) doi:10.1016/j.regsciurbeco.2010.03.003. The estimator maximizes the full log-likelihood function in which the parameter of spatial dependence is constrained.
Usage
mml(Rho, ff, dataset, wmat, var.agg, m = 10)
Arguments
Rho |
the constrained parameter of spatial dependence |
ff |
Formula of the linear model. It excludes the spatial lag |
dataset |
Data frame with the data |
wmat |
Spatial weight matrix |
var.agg |
Spatial index of the data frame |
m |
How many time periods have passed since the beginning of the space-time process |
Value
The estimates tables
Examples
set.seed(123)
sd = sim_data_fe(dataset=regsamp,N=50,TT=6,
spatial = 80,Tau = -0.2,Rho = 0.4,
Beta = 2,sdDev = 2,startingT = 10,
LONGLAT = TRUE);sd[[1]]$X2 = stats::rnorm(nrow(sd[[1]]@data))
est_mml = mml(dataset = sd[[1]]@data,Rho = 0.4,
ff = Y~X1+X2,
wmat = sd[[2]],var.agg = c('Anno','Cod_Provincia'),
m = 10)
est_mml
Cropland grid of Italy (100km x 100km squares)
Description
SpatialPolygonsDataFrame object with croplands of Italy approximated with 100km x 100km squared polygons. It is based on croplands data contained in the Corine Land Cover 2012 raster map.
Usage
r100km
Format
SpatialPolygonsDataFrame object
Source
https://land.copernicus.eu/pan-european/corine-land-cover/clc-2012
Cropland grid of Italy (20km x 20km squares)
Description
SpatialPolygonsDataFrame object with croplands of Italy approximated with 20km x 20km squared polygons. It is based on croplands data contained in the Corine Land Cover 2012 raster map.
Usage
r20km
Format
SpatialPolygonsDataFrame object
Source
https://land.copernicus.eu/pan-european/corine-land-cover/clc-2012
Cropland grid of Italy (40km x 40km squares)
Description
SpatialPolygonsDataFrame object with croplands of Italy approximated with 40km x 40km squared polygons. It is based on croplands data contained in the Corine Land Cover 2012 raster map.
Usage
r40km
Format
SpatialPolygonsDataFrame object
Source
https://land.copernicus.eu/pan-european/corine-land-cover/clc-2012
Cropland grid of Italy (60km x 60km squares)
Description
SpatialPolygonsDataFrame object with croplands of Italy approximated with 60km x 60km squared polygons. It is based on croplands data contained in the Corine Land Cover 2012 raster map.
Usage
r60km
Format
SpatialPolygonsDataFrame object
Source
https://land.copernicus.eu/pan-european/corine-land-cover/clc-2012
Coordinates of simulated farms in Northern Italy
Description
SpatialPointsDataFrame object with 1000 SpatialPoints to represent simulated farms in the simulation exercise
Usage
regsamp
Format
SpatialPointsDataFrame object
Merged NUTS3 for simulation exercise
Description
SpatialPolygons object with merged NUTS3 regions of Northern Italy used in the simulation exercise
Usage
sel_regioni
Format
SpatialPolygons object
Source
ISTAT
Simulate space-time stochastic process with fixed-effect
Description
This function simulates a space-time stochastic process according to the defined spatial structure and input paramters. It simulates data of a dynamic spatial lag model. It includes one exogenous variable and a fixed-effect correlated with the exogenous variable.
Usage
sim_data_fe(
dataset,
N,
TT,
spatial = 100,
Tau = -0.14,
Rho = 0.67,
Beta = 1,
sdDev = 5,
startingT = 11,
LONGLAT = TRUE
)
Arguments
dataset |
SpatialObject with the spatial units for which the data will be simulated |
N |
How many spatial units will be used |
TT |
Time dimension of the simulated process |
spatial |
Radius that defines the scope of spatial dependence |
Tau |
Autocorrelation parameter |
Rho |
Spatial dependence parameter |
Beta |
Coefficient associated to the exogenous variable |
sdDev |
Standard Deviation of the (gaussian) error term |
startingT |
The number of time periods after which the simulated data will be recorded |
LONGLAT |
Boolean. If the projection is longlat |
Value
A list with two objects. The first object is the STFDF with the simulated data. The second object is the spatial weight matrix
Examples
library(spacetime)
library(sp)
library(spdep)
set.seed(123)
sd = sim_data_fe(dataset=regsamp,N=100,TT=8,
spatial = 80,Tau = -0.2,Rho = 0.4,
Beta = 2,sdDev = 2,startingT = 10,
LONGLAT = TRUE)
stplot(sd[[1]][,,'Y'])
dev.new()
plot(sel_regioni)
points(coordinates(sd[[1]]@sp))
plot(mat2listw(sd[[2]]),coordinates(sd[[1]]@sp),add=TRUE,col=2)
Neighbors of NUTS3 of Sardinia
Description
A table containing the link of NUTS3 in Sardinia with NUTS3 of mainland Italy
Usage
tabSard
Format
A data frame with the links