version 0.3-8

ChangeLog

@@ -526,3 +526,14 @@ processing col and pch arguments
 2018-08-27  Andreas Papritz  <papritz@env.ethz.ch>
 
 * georob.S3methods.R (add1.georob, drop1.georob, step.georob) elimination of data argument
+
+
+2018-11-25 Andreas Papritz  <papritz@env.ethz.ch>
+
+* georob.simulation.R (f.aux.sim.1) elimination of grid.nodes argument
+
+
+2019-03-29 Andreas Papritz  <papritz@env.ethz.ch>
+
+* georob.private.functions.R (f.aux.gcr) call to RFvariogram{RandomFields, version 3.1} replaced by call to RFfctn{RandomFields, version 3.3}
+* georob.simulation.R (condsim) conditioning data is passed as SpatialPointsDataFrame to RFsimulate 

DESCRIPTION

@@ -1,15 +1,15 @@
 Package: georob
 Type: Package
 Title: Robust Geostatistical Analysis of Spatial Data
-Version: 0.3-7
-Date: 2018-08-27
+Version: 0.3-8
+Date: 2019-03-29
 Authors@R: c(
   person( "Andreas", "Papritz", role = c( "cre", "aut" ), 
           email =  "andreas.papritz@env.ethz.ch" ),
   person( "Cornelia", "Schwierz", role = "ctb" ))
 Depends: R(>= 2.14.0), parallel, snowfall, sp(>= 0.9-60), stats
 Imports: abind, constrainedKriging(>= 0.2-1), fields, lmtest, methods,
-        nlme, nleqslv, quantreg, RandomFields(>= 3.0.10), robustbase(>=
+        nlme, nleqslv, quantreg, RandomFields(>= 3.3.6), robustbase(>=
         0.90-2)
 Suggests: geoR, gstat, multcomp, lattice
 Description: Provides functions for efficiently fitting linear models with 
@@ -22,9 +22,9 @@ Description: Provides functions for efficiently fitting linear models with
          Kriging predictions of log-transformed data.
 License: GPL (>= 2)
 NeedsCompilation: no
-Packaged: 2018-08-27 18:21:24 UTC; papritz
+Packaged: 2019-03-29 15:48:40 UTC; papritz
 Author: Andreas Papritz [cre, aut],
   Cornelia Schwierz [ctb]
 Maintainer: Andreas Papritz <andreas.papritz@env.ethz.ch>
 Repository: CRAN
-Date/Publication: 2018-08-27 19:02:18 UTC
+Date/Publication: 2019-03-29 16:10:02 UTC

MD5

@@ -1,30 +1,30 @@
-522ba0488556b0fda02a63a13fc5c36a *ChangeLog
-7ca58ea0d9abf49632f47c5e74f20f53 *DESCRIPTION
-c495d68b9fa7cd72e5c65e9028a77b12 *NAMESPACE
-0125da1be2304e5f9db0bb28881d63db *NEWS
+8bfcbd8d7d5cdf20e8644956710f2b62 *ChangeLog
+79d0d3ca67efc985b4c566cb2dccaade *DESCRIPTION
+1bbbbd8d10c38e8072e95e6ec7124bed *NAMESPACE
+4b14bc84569226fff150e55321f1c6e1 *NEWS
 cf3b31b5c1d9a2523c5b3fad53756e19 *R/georob.S3methods.R
 690b7f43f59cff872ae22dba73b4e48b *R/georob.cv.R
 f1d1f5a5adb4a085fe41b4d9faace95f *R/georob.exported.functions.R
 be29537ab86336681a92df97bbb94023 *R/georob.lgnpp.R
-71c52f799f30714f5ac4148820b215a3 *R/georob.predict.R
-a4da70c241c5dd02ce9e75a4f6787a2e *R/georob.private.functions.R
-53aa2ba7843ce913d4509ef128afacb2 *R/georob.simulation.R
+869eb984d6fc82469fb194ecf74453b0 *R/georob.predict.R
+6183183fdf66cae67036b5017c5eb484 *R/georob.private.functions.R
+d0e7e17d6c095fcbfdb8fc6f32310c7d *R/georob.simulation.R
 fe99bbce2cb803b386fb1c916283e6f8 *R/variogram.R
-018dbdc6e47c8e326ed72cf4735ae4d2 *build/vignette.rds
+59ab64eb12c566599f6eef932e6605ff *build/vignette.rds
 831ce9ec4e36d929c6a6fa482f3121db *inst/doc/georob_vignette.R
-8ab0a6ade690bbdefe0f2b0d6bb37d12 *inst/doc/georob_vignette.Rnw
-ccfd017b418ad6a50c9b8bab92b01b4e *inst/doc/georob_vignette.pdf
+a30a7d6fd40888e72c53bc8f30d8ecb3 *inst/doc/georob_vignette.Rnw
+145056aae06385afc9a08bbad606d41b *inst/doc/georob_vignette.pdf
 2f3793f9b337bf1e73d2df9de6d1d01e *man/compress.Rd
 69deb3505141d34494cf09906a169192 *man/control.georob.Rd
 7f7a491250c86813ddbd3c174b71f2b4 *man/cv.Rd
 da8aade8683a3498bc59a82c9823627c *man/cv.georob.Rd
 58ad2f1d46122ba59ebd41680d69bd37 *man/fit.variogram.model.Rd
-2c8779efffd16249840853701db29c08 *man/georob.Rd
+7c3b5b7d94fb144bdca56a0d77978397 *man/georob.Rd
 40e7c3f64dd94b09179f7f84a6bbd698 *man/georobModelBuilding.Rd
 4f76714367b5419d4f0feae5b2d8de64 *man/georobObject.Rd
 6f561758aef26ff0bae01145a2500293 *man/georobPackage.Rd
 bee9ac39f7bae2274ad1eb918c770ad1 *man/georobS3methods.Rd
-6b84596d9cfaea3b867c7d786f8cdb4b *man/georobSimulation.Rd
+0d5b7a61302f4883503a7979572f07d7 *man/georobSimulation.Rd
 bfeffa190c881a7d1559976cce557e7c *man/internal.functions.Rd
 020482848aaf4399f6d5187bcfc421cd *man/lgnpp.Rd
 6b237a1a67a76099dc74f6d9248caf60 *man/param.defaults.Rd
@@ -38,7 +38,7 @@ c6c6d5e6ee70c8c1a3ec26aa9ba7815b *man/validate.predictions.Rd
 bb27c0989db96d417b6420d1b941bf53 *vignettes/0-block-discretization.pdf
 43e94e7c726e507b341dc8dbbab2c3c3 *vignettes/0-overview-zinc-meuse.pdf
 fcb9c9341d601d8ed6e6be5066b5c0ba *vignettes/georob.bib
-8ab0a6ade690bbdefe0f2b0d6bb37d12 *vignettes/georob_vignette.Rnw
+a30a7d6fd40888e72c53bc8f30d8ecb3 *vignettes/georob_vignette.Rnw
 ef97150509737b3d7c819a8f5d597538 *vignettes/natbib.bst
-5f4e209426d612410a9c56f0618abadb *vignettes/r_coalash_objects.RData
-094845e7a8adff61418d2dcf0ed33136 *vignettes/r_meuse_zinc_objects.RData
+d0a78cfd6a261091b545dd3b0a84c449 *vignettes/r_coalash_objects.RData
+978816df451f639c514f19724bf7782e *vignettes/r_meuse_zinc_objects.RData

NAMESPACE

@@ -10,7 +10,7 @@ importFrom( methods, slot)
 importFrom( nlme, fixef, fixed.effects, ranef, random.effects )
 importFrom( nleqslv, nleqslv )
 importFrom( quantreg, rq.fit )
-importFrom( RandomFields, RFsimulate, RFvariogram, RFoldstyle, RFoptions)
+importFrom( RandomFields, RFsimulate, RFfctn, RFoldstyle, RFoptions)
 importFrom( robustbase, lmrob.control, lmrob.fit, Qn, robMD, summarizeRobWeights )
 # importFrom( nor1mix, norMix, pnorMix, qnorMix, var.norMix )
 

NEWS

@@ -14,4 +14,6 @@
 2017-05-09  georob Version 0.3-4 released to CRAN, correction of minor bug for computing robust initial values of regression coefficients by rq().
 2017-10-20  georob Version 0.3-5 released to CRAN, new function coef().
 2018-01-26  georob Version 0.3-6 released to CRAN, new function condsim() for (un)conditional simulation from a fitted georob object.
-2018-08-27  georob Version 0.3-6 released to CRAN, minor change of arguments of functions for model building.
+2018-08-27  georob Version 0.3-6 released to CRAN, minor change of arguments of functions for model building.
+2018-11-25  georob Version 0.3-7 released to CRAN, minor change of function for (conditional) simulation.
+2019-03-29  georob Version 0.3-8 released to CRAN, changes required by new version 3.3.6 of package RandomFields

R/georob.predict.R

@@ -102,7 +102,7 @@ function(
   ## setup or check contents of variogram.object
 
   if( !all( 
-      is.null(variogram.model), is.null(param),  is.null(aniso), 
+      is.null(variogram.model), is.null(param), is.null(aniso), 
       is.null(variogram.object)
     )
   ){

R/georob.private.functions.R

@@ -754,6 +754,8 @@ function(
   ## 2016-08-09 AP changes for nested variogram models
   ## 2016-08-10 AP changes for isotropic variogram models
   ## 2017-12-23 AP improved memory management in parallel computations
+  ## 2019-03-29 AP call to RFvariogram{RandomFields, version 3.1} replaced by 
+  ##               call to RFfctn{RandomFields, version 3.3}
 
   ## check consistency of arguments
 
@@ -765,8 +767,6 @@ function(
     gcr.constant <- as.list( rep( NA_real_, length(variogram.object) ) )
   }
 
-
-
   res <- lapply(
     1L:length(variogram.object),
     function( i, x, gcr.constant, lag.vectors, control.pcmp ){
@@ -805,14 +805,34 @@ function(
 
           ## objects s, e, lag.vectors, model.list taken from parent environment
 
+          #           RandomFields Version 3.1
+          #           result <- try(
+          #             -RFvariogram(
+          #               x = lag.vectors[s[i]:e[i], ], model = model.list, 
+          #               dim = attr( lag.vectors, "ndim.coords" ), grid = FALSE
+          #             ),
+          #             silent = TRUE
+          #           )
+
+          RFoptions( allow_duplicated_locations = TRUE )
+
+          ## note: RFfctn computes covariance for stationary and negative
+          ## semivariance for IRF models, required is negative semivariance
           result <- try(
-            -RFvariogram(
+            RFfctn(
               x = lag.vectors[s[i]:e[i], ], model = model.list, 
               dim = attr( lag.vectors, "ndim.coords" ), grid = FALSE
-            ),
+            ) - model.list[["var"]],
             silent = TRUE
           )
+
+          RFoptions( allow_duplicated_locations = FALSE )
+
           if( !(identical( class( result ), "try-error" ) || any( is.na( result ) )) ){
+            if(!variogram.model %in% irf.models){
+              ## subtract variance for stationary models
+              result <- result - model.list[["var"]]
+            }
             result
           } else {
             "RFvariogram.error"
@@ -843,14 +863,34 @@ function(
 
           ## objects s, e, lag.vectors, model.list taken from parent environment
 
+          #           RandomFields Version 3.1
+          #           result <- try(
+          #             -RFvariogram(
+          #               x = lag.vectors[s[i]:e[i]], model = model.list, 
+          #               grid = FALSE 
+          #             ),
+          #             silent = TRUE
+          #           )
+
+          RFoptions( allow_duplicated_locations = TRUE )
+
+          ## note: RFfctn computes covariance for stationary and negative
+          ## semivariance for IRF models, required is negative semivariance
           result <- try(
-            -RFvariogram(
+            RFfctn(
               x = lag.vectors[s[i]:e[i]], model = model.list, 
               grid = FALSE 
             ),
             silent = TRUE
           )
+
+          RFoptions( allow_duplicated_locations = FALSE )
+
           if( !(identical( class( result ), "try-error" ) || any( is.na( result ) )) ){
+            if(!variogram.model %in% irf.models){
+              ## subtract variance for stationary models
+              result <- result - model.list[["var"]]
+            }
             result
           } else {
             "RFvariogram.error"
@@ -911,7 +951,7 @@ function(
 
       not.ok <- any( sapply( 
           Valpha, 
-          function( x ) identical( x, "RFvariogram.error" ) || is.na(x)
+          function( x ) identical( x, "RFvariogram.error" ) || any(is.na(x))
         ))
 
       if( !not.ok ){

R/georob.simulation.R

@@ -50,6 +50,10 @@ condsim <- function(
   ## parameters estimated by georob for a spatial linear model
 
   ## 2018-01-27 A. Papritz
+  ## 2018-11-25	A. Papritz small change in function f.aux.sim.1
+  ## 2019-03-29 A. Papritz restore default for RFoption spConform
+  ## 2019-03-29 A. Papritz conditioning data is passed as SpatialPointsDataFrame to
+  ##                       RFsimulate (only applies for exact coordinates)
 
 
   ## auxiliary function for aggregating a response variable x for the
@@ -544,7 +548,6 @@ condsim <- function(
 
   }
 
-
   ## compute (conditional) realizations ...
 
   if( !control[["use.grid"]] ){
@@ -592,6 +595,7 @@ condsim <- function(
 
     if( control[["condsim"]] ){
       data <- cbind( coords.d, values = rep( 0., NROW(coords.d) ) )
+      coordinates(data) <- colnames(coords.d)
     } else {
       data <- NULL
     }
@@ -612,7 +616,7 @@ condsim <- function(
       tmp <- lapply(
         variogram,
 
-        function( x, grid.nodes, nsim ){
+        function( x, nsim ){
 
           signal <- RFsimulate(
             model = x[["signal"]], n = nsim,
@@ -630,7 +634,7 @@ condsim <- function(
             signal
           }
 
-        }, grid.nodes = grid.nodes, nsim = nsim
+        }, nsim = nsim
       )
 
       res <- tmp[[1]]
@@ -640,6 +644,8 @@ condsim <- function(
         }
       }
 
+      RFoptions( spConform = TRUE )      
+
       res
 
     }
@@ -842,6 +848,8 @@ condsim <- function(
         }
       }
 
+      RFoptions( spConform = TRUE )
+
       res
 
     }

inst/doc/georob_vignette.Rnw

@@ -12,8 +12,9 @@
 % 2016-11-14 AP Korrektur Fehler in Rotationsmatrix f\"{u}r geom. anisotrope Variogramme
 % 2017-08-11 AP Korrektur von Tippfehlern
 % 2017-10-20 AP Korrektur von Fehler bei Plot von Profillikelihood
+% 2018-11-16 AP Kleine Erg\"{a}nzung Abschnitt \"{u}ber Berechnung von robusten Startwerten
 
-% cd R.user.home/georob/vignettes/
+% cd ~/R.user.home/georob/vignettes/
 % R CMD Sweave georob_vignette.Rnw;  R CMD Stangle georob_vignette.Rnw; pdflatex georob_vignette.tex; open georob_vignette.pdf;
 % bibtex georob_vignette; pdflatex georob_vignette.tex;  open georob_vignette.pdf;
 
@@ -833,7 +834,7 @@ We now study the \code{summary} output in detail:
 % log-likelihood is equal to -54.6.
 %
 The estimated variogram parameters are reported along with 95\%
-confidence intervals that are computed based on the asympotic normal
+confidence intervals that are computed based on the asymptotic normal
 distribution of (RE)ML estimates from the observed Fisher information.
 
 The dependence of \code{log(zinc)} on \code{sqrt(dist)} is highly
@@ -871,7 +872,7 @@ waldtest(r.georob.m0.spher.reml, .~.+soil)
 
 \smallskip
 
-Drift models may also be build by stepwise covariate selection based
+Drift models may also be build by step-wise covariate selection based
 on AIC, either keeping the variogram parameters fixed (default)
 %
 <<meuse-zinc-georob-reml-step-1,eval=FALSE>>=
@@ -933,7 +934,7 @@ Variogram:  RMspheric
 \end{Soutput}
 \end{Schunk}
 %
-which selects the same model.  Note that stepwise covariate
+which selects the same model.  Note that step-wise covariate
 selection by \code{step.georob()}, \code{add1.georob()} and
 \code{drop1.georob()} requires the \emph{non-restricted}
 log-likelihood.  Before evaluating candidate models, the initial model
@@ -1091,7 +1092,7 @@ plot(upr, position=c(2/3, 0, 1, 1), more=FALSE)
 @ 
 %
 \ecf{Lognormal point Kriging prediction of \code{zinc} (left:
-prediction; middle and right: lower and upper bounds of pointwise 95\%
+prediction; middle and right: lower and upper bounds of point-wise 95\%
 prediction intervals).}{meuse-zinc-point-Kriging-plot}
 %
 
@@ -1164,7 +1165,7 @@ plot(upr, position=c(2/3, 0, 1, 1), more=FALSE)
 %
 \ecf{Lognormal block Kriging prediction of \code{zinc} computed under
 the assumption of permanence of log-normality (left: prediction; middle
-and right: lower and upper bounds of pointwise 95\% prediction
+and right: lower and upper bounds of point-wise 95\% prediction
 intervals).}{meuse-zinc-block-Kriging-plot}
 %
 
@@ -1173,7 +1174,7 @@ log-normal laws~---~which strictly cannot hold~---~does not much
 impair the efficiency of the back-transformation as long as the blocks
 are small \citep{Cressie-2006,Hofer-etal-2013}.  However, for
 larger blocks, one should use the optimal predictor obtained by
-\emph{averaging back-transtransformed point predictions}.  \code{lgnpp()}
+\emph{averaging back-transformed point predictions}.  \code{lgnpp()}
 allows to compute this as well.  We illustrate this here by predicting
 the spatial mean over two larger square blocks:
 %
@@ -2376,9 +2377,10 @@ of initial values than maximizing the Gaussian (restricted)
 log-likelihood with respect to the same parameters.  If the initial
 values for \code{param} and \code{aniso} are not sufficiently close to
 the roots of the system of nonlinear equations, then \code{nleqslv()}
-may fail to find them.  Setting \code{initial.param = TRUE} allows one
-to find initial values that are often sufficiently close to the roots
-so that \code{nleqslv()} converges.  This is achieved by:
+may fail to find them.  Setting \code{initial.param = TRUE} (see
+\code{control.georob()}) allows one to find initial values that are
+often sufficiently close to the roots so that \code{nleqslv()}
+converges.  This is achieved by:
 
 \begin{enumerate}
 

man/georob.Rd

@@ -1,4 +1,4 @@
-% 2015-11-27 A. Papritz
+% 2018-11-16 A. Papritz
 % R CMD Rdconv -t html -o bla.html georob.Rd ; open bla.html; R CMD Rd2pdf --force georob.Rd; 
 
 \encoding{macintosh}
@@ -356,8 +356,8 @@ georob(formula, data, subset, weights, na.action, model = TRUE,
     values for \code{param} and \code{aniso} are not sufficiently close to
     the roots of the system of nonlinear equations, then
     \code{\link[nleqslv]{nleqslv}} may fail to find them.      
-    Setting
-    \code{initial.param = TRUE} allows one to find initial values that are
+    Setting \code{initial.param = TRUE} (see \code{\link{control.georob}}) 
+    allows one to find initial values that are
     often sufficiently close to the roots so that
     \code{\link[nleqslv]{nleqslv}} converges.  This is achieved by: