		=============================
		CHANGES IN geoR VERSION 1.3-7
		=============================

NEW FEATURES
	
     o changes in as.geodata to keep factor attributes of the covariates

     o type of contrast used is now returned in likfit

     o new data-sets "ca20" and "hoef" included
 


BUGS

     o varcov.spatial now tries the svd() decomposition when chol() fails. 
   
     o likfit issues messages that parameter estimation for the "power" and 
       "gneiting.matern" models is not implemented.

     o change in the measure of variability in backtransform.moments.
       (q0.75 - q0.25)/4 replaced by (q0.975 - q0.025)/4
       and the restult is calles uncertainty instead of variability

     o bug fix in variog.mc.env when lambda != 1 and for directional variograms

     o bug fix in proflik when fix.lambda = FALSE



		=============================
		CHANGES IN geoR VERSION 1.3-6
		=============================

NEW FEATURES
	
     o improvements in the documentation for output.control     
       with better explanations about the option signal
 
     o changes in image.grf now allowing inclusion of legends
       and better control of graphical parameters

BUGS
   
     o fix in kriging variances in krige.conv 


				
		=============================
		CHANGES IN geoR VERSION 1.3-5
		=============================

NEW FEATURES

     o improvements in odcumentation of output.control     
       with better explanations about the option signal
 
     o changes in the axis limits in image.grf() 


BUGS
   
     o krige.bayes and krige.conv:
       bug fix when simulation are required with nugget > 0 
       and prediction locations coincident with data locations.

     o boxcox.fit() now uses optimise() for one dimension minimisation

     o proflik() now uses optimise() for one dimension minimisations

     o bug fix in plot.geodata() whne argument x is missing.

     o as.geodata() now aways returns element covariates as
       a data-frame when covariates are included in the model.


				
		=============================
		CHANGES IN geoR VERSION 1.3-3
		=============================

NEW FEATURES

     o as.geodata() tests for NA's in the coordinates
     
     o likfit() now uses optimize() instead of optim() when
       minimization is unidimensional  

     o  tests for size of the trend matrix in comparison
        with the number of coordinates
				
     o  improvements in trend.spatial() and its documentation

BUGS
   
     o bug fix in varcov.spatial() causing craches in repetitive uses of this
       function, as for example when called by likfit() in simulation studies.

     o bug fix and modifications in variofit() for covariance models  
	"linear" and "power". 

     o bug fixed for the power model in cov.spatial and functions in geoR.c
			


		=============================
		CHANGES IN geoR VERSION 1.3-2
		=============================

NEW FEATURES
   
     o changes in the documentation of Box-Cox related functions

     o trend.spatial now assigns a class "trend.spatial" to its output

     o functions with argument trend can now also take an output of the 
       function trend.spatial()

     o changes in calls to C code
        

BUG FIXES

     o fix in variog.model.env() and variog.mc.env() when argument "trend" is used

     o bug fix in krige.bayes when argument messa=F

     o bug fix in image.kriging() and image.krige.bayes()

     o fix and additions in the documentation of trend.spatial()

     o bug fix in print.grf(), krige.bayes.grf() 
       caused by changes in version 1.3-0
 
     o in plot.xvalid() when "borders = NULL"

     o more details on the argument realisations in documentation for
       likfit and as.geodata



		=============================
		CHANGES IN geoR VERSION 1.3-0
		=============================

NEW FEATURES
   
     o the file geoRintro.pdf is no longer included in the package 
       distribution due to its size.
       The file remain available on the web on html, ps and pdf formats

     o lines.variomodel() is now a generic functions.
       and the previous lines.variomodel() is moved to
       lines.variomodel.default().
       So far methods implemented for lines.variomodel.krige.bayes(),
       lines.variomodel.grf(), lines.variomodel.likGRF() and
       lines.variomodel.variofit()

     o changes in lines.krige.bayes():
       - lines.krige.bayes moved to lines.variomodel.krige.bayes().
       - lines.variomodel.krige.bayes() now computes and plots 
         summaries of the posterior for the variogram functions
         as the default option. 
       - The previous functionality using summaries of the posterior 
          of the model parameters remains available as non-default.

     o 	read.geodata() and as.geodata() have now options to deal 
        with more that one data at the same location

     o 	read.geodata() and as.geodata() have now options to deal 
        with NA's

     o 	read.geodata() and as.geodata() returns $m.var, 
        the measurement variance when there is more than 
        one data at the same location


BUG FIXES
 
     o AIC and BIC in likfit() now computed as:
            AIC = 2log(L) + 2p
            BIC = 2log(L) + p log(n)

     o bug fix in plot.proflik() when 2D profiles not computed

     o variog() now returns $max.dist which is used as default option in variofit()

     o bug fix in weigths $n variofit() when max.dist was provided by the user

     o documentation of variog(): fix in the equation of the variograms estimators

     o bug fixes in legend.krige()



		=============================
		CHANGES IN geoR VERSION 1.2-8
		=============================

NEW FEATURES
   
     o as.geodata() now saves stores the covariates in a data.frame format
	
     o changes in grf() and related functions to facilitate simulations in 1-D

     o bug fix in persp.grf()

     o change in argument name in plot.grf(): plot.grid changed to plot.locations 

     o image.grf() and persp.grf() now recognises simulations in 1-D and produces a 
       2D plot to shown the simulations (intead of an image or perspective plot)

     o as.geodata() force numerical values in the $coords and $data outputs 

     o change in the arrangement of plots in plot.geodata() 



BUG FIXES

     o fixes/changes in the documentation, geoRintro and geoR web-pages

     o fix in trend.spatial() message

     o fix in coords argument in plot.geodata(), which now also accepts an
       object of the class "geodata"



		=============================
		CHANGES IN geoR VERSION 1.2-5
		=============================

NEW FEATURES
   
     o more details on the loss function in the documentation of variofit


BUG FIXES

     o bug fix in calculation of directional variograms in variog()

     o bug fix in krige.bayes documentation

     o fix in the documentation of internal and defunct functions.

     o fix in test for NA's to conform with R-1.5.0

     o replacement of T/F by TRUE/FALSE



		=============================
		CHANGES IN geoR VERSION 1.2-1
		=============================

NEW FEATURES

     o some tests for matrix inversions when the model has  
       covariates with very different orders of magnitude 
  
     o outputs of prior.control(), model.control(), output.control()
       and krige.control() have now classes "prior.geoR", "model.geoR",
       "output.geoR", "krige.geoR", respectivelly. 

     o few checks moved from krige.conv to krige.control

     o argument "realisation" in as.geodata can now also 
       takes a column number or a name of an element of the list
       geodata.

     o the arguments model, prior and output in krige.bayes
       can now take not only calls to the respective control functions,
       but also lists as usual for control functions. Partial matching
       for arguments is implemented.

     o idem previous for the arguments krige and output in the
       function krige.conv.

     o futher checks for the validity covariance matrix in prior.control
       when using a normal prior for the parameter beta
 
     o the data set s256i from the Tech. Rep. which describes the 
       implementation of krige.bayes is now included 

     o futher checks in rinvchisq


BUG FIXES

     o bug fix when using trend different from constant.
       This bug was introduced in version 1.2-0

     o bug fix in likfit when using multiple initial values
       This bug was introduced in version 1.2-0

     o several corrections in the documentation 



		=============================
		CHANGES IN geoR VERSION 1.2-0
		=============================

NEW FEATURES

     o krige.bayes() now allows for "user defined" (discrete) prior
       distributions for the parameters phi and tausq.rel

     o new function post2prior() which prepares the posterior
       obtained in a run of krige.bayes to be used as a prior
       in a subsequent run

     o function likfit() now allows data assumed to be from
       different and independent realisations.
       It has now an extra argument "realisations" which
       takes a vector of indicating the number of the realisation.

     o functions read.geodata() and as.geodata() also includes
       the argument realisations

     o function trend.spatial changed to accept different ways to
       read covariates in the model
      
     o plot.variogram has an extra argument which allows points
       sizes to be proportional to the (square root) of the number
       of pairs in each bin


BUG FIXES

    o  bug fix in picking tausq.res values in sample.posterior()

    o  bug fix in lines.variomodel() for scaled variogram
    
    o  likfit() with estimated lambda and RML does not crashes anymore


		=============================
		CHANGES IN geoR VERSION 1.1-5
		=============================

NEW FEATURES

     o function points.geodata(): argument 'pt.sizes' changed to 'pt.divide' 
       internal back compatibility included
 
     o argument 'pt.divide' can now also take a numerical vector of quantiles
       to divide the data points in classes	
 
     o plot.xvalid has more options for selecting plots

     o plot.variogram passes more arguments to the plotting function
       when ploting multiple variograms


BUG FIXES

    o  fix in variog() to allow computation with points at coincident locations
    
    o  as.geodata now returns covariates as matrix instead of data-frame
       to prevent problems in formulas

    o  variog() now checks for incompatible dimentions of coords 
       and data

    o  fixes in the documentation following checks for R-1.4.0



		=============================
		CHANGES IN geoR VERSION 1.1-2
		=============================


NEW FEATURES

    o  new function hist.krige.bayes() which plots histograms
       and/or density estimation of the samples from the posterior
       of the model parameters

    o  new function plot.krige.bayes() plots discrete prior and posterior
       for the parameters phi and/or tausq.rel

    o  BoxCox group of functions including parameter estimation 
       by maximum likelihood, random generation, density
       and methods for printing and ploting

    o  krige.bayes, krige.conv and ksline now selects prediction
       locations inside borders. Selection computed by the 
       new auxiliary function locations.inside()

    o  plot.xvalid now reads attributes from output of xvalid 

    o  internal changes in C code

    o  image.kriging, image.krige.bayes, persp.kriging and 
       persp.krige.bayes now reads attributes of the related function 
	
    o  the version number is now printed when geoR is loaded

    o  Windows distribution now includes .chtml help files

    o  Matern correlation function now implemented in the  
       .c code "cor_diag" 
    
    o  new C wrapper function corr.diaglowertri for cor_diag

    o  major reformulation of the function krige.bayes including
   
       - "conjugate" priors for beta and sigmasq

       - flexibility in which parameters are considered random
 
       - unified and more versatile output for whatever prior 
         specification

       - parameters are reffered by greek letters and not
         using geostatistical jargon

       - internal changes to improve efficiency and memory handling

       - changes in arguments of control functions

       - inclusion of prior information in the output

       - major reformulation in the output of krige.bayes()

    o  new function sample.prior() to sample from priors used in 
       krige.bayes()

    o  new function sample.posterior() to sample from posteriors 
       returned by krige.bayes()

    o  new function statistics.predictive() to compute statistics
       from simulations of predictive returned by krige.bayes()
       and/or krige.conv()

    o  grf returns "spacing" or "xspacing" and "yspacing"
       attributes for regular grids

    o  output.control is an isolated function with its own documentation
       and now also used by krige.conv() 

    o  krige.conv() can return outputs from the posterior distribution
       as for krige.bayes. This includes quantiles, thresholds
       and others defined by output.control()

    o  new functions for the Scaled-Inverse-ChiSquare distribution
       rinvchisq() and dinvchisq

    o krige.bayes() and krige.conv() now backtransform moments analitically
      for kappa = 0.5

    o  Functions (typically) internally used

    	o  new function cond.sim() to simulate from predictive.
           Typically internally called by krige.conv() and krige.bayes()

    	o  new function print.betavar() for nive print of covariance matrices

    	o  new function beta.sigmasq.post() for key calculations on 
           posteriors and predictive

        o new functions transform.data() and backtransformed.moemnts()


BUG FIXES
    
    o  parentesis matching in documentation for krige.conv()
       and xvalid() 

    o variance in krige.bayes for normal prior for the parameter beta



		=============================
		CHANGES IN geoR VERSION 1.0-8
		=============================


NEW FEATURES

    o  new function legend.krige() to add legend to
       image plots

    o  arguments in image.kriging() and image.krige.bayes()
       to include legend in the plot

    o  argument "obj.model" included in krige.conv()

    o  argument "borders" included in plot.geodata()

    o  argument "lambda", "trend" and "weights.divide"
       included in points.geodata()


BUG FIXES

    o  fix in variog which was causing problems in NA handling for variog4()

    o  a few bug fixes in proflik()

    o  argument pch in points.geodata()

    o  constant part of the likelihood included in the trace of 
       likfit() when print.pars = TRUE

    o  argument "trend" in plot.geodata() now working properly




		=============================
		CHANGES IN geoR VERSION 1.0-7
		=============================


NEW FEATURES

    o  new argument "borders" for image.kriging(), persp.kriging(),
       image.krige.bayes(), persp.krige.bayes().

    o  function points.geodata() takes argument "borders" to include borders
       of the region in the plot.

    o  additional optional argument in plot.xvalid() for including 
       the borders of the area in the errors plot

    o  argument "ask" included in plot.proflik()

    o  lines.variomodel() and lines.krige.bayes() 
       now uses curve() to draw the variogram model. 
       Some arguments have been changed.


BUG FIXES

    o  fixed bug when using option scale=T in lines.variomodel.
       Internal changes in this functions.

    o  proflik() missing jacobian value for $\lambda \neq 1$ and fixed 

    o  problem with splines and duplicated values in plot.proflik() 

    o  variog4 and plot.variog4: NA's and different lags

    o  krige.bayes: changes in arguments names of nugget parameters
       to make clear when relative nugget is used.
       
    o  repairs in the documentation of krige.bayes()
    
    o  fixed bug in xvalid when entering with external validation data




		=============================
		CHANGES IN geoR VERSION 1.0-6
		=============================

NEW FEATURES

    o   function geoRdefunct() lists defunct functions

    o   likfit.old(), olsfit() and wlsfit() are now defunct

    o   this file is included in the package distribution at 
        .../geoR/docs/NEWS

    o	changes in the variogram based parameter estimation

        - new function variofit() for variogram based estimation via
          ordinary or weighted least squares replaces olsfit() and wlsfit()

	- parameter kappa, always fixed in wlsfit() and olsfit()
          can now be estimated in by variofit()

	- print.variomodel() and summary.variomodel() changed accordingly 


    o  file random.c no longer user. R's random number generator 
       is now used in file woodandchan.c


    o  flexiblity to change graphical parameter is plot.variog4()


    o  xvalid() incorporates prediction at differet locations


    o  extra graphics in plot.xvalid()


    o data(wolfcamp) now available


	
BUG FIXES

    o  likfit() : default optiom fir likfit.limits() when 
       cov.model = "powered.exponential"




		  =============================
		  CHANGES IN geoR VERSION 1.0-5
		  =============================

NEW FEATURES

    o	directional variograms

        -  variog() now computes directional variograms using new arguments
           "direction" and "tolerance"

	-  variog4() computes variograms for four directions

	-  plot.variog4() plots output of variog4()


    o  file random.c no longer user. R's random number generator 
       is now used in file woodandchan.c


    o  cross-validation

        - new function xvalid() to perform cross-validation

	- associated plot method: plot.xvalid()


    o  power model for covariance functions implemented


    o  read.geodata() warns if data at coincident locations

  
	
BUG FIXES

    o  several argument matching fixes in the documentation

    o  print.summary.likGRF() now works properly

    o  argument fixing on plot.grf() 

    o  .First.lib() following R's standards and source moved to zzz.R

    o  causes of warning messages in wlsfit() and olsfit() were removed 

    o  error fixing in points.geodata() "pt.size not found"



	**************************************************
	*						 *
	*	 geoR 1.0-0 : RELEASE STATMENT	         *
	*						 *
	**************************************************


  What's geoR:
  -------------
    geoR is a package to perform geostatistical data analysis and 
    spatial prediction, expanding the tools already available in 
    R for analysis of spatial data.

  Legalese:
  ---------
  geoR is distributed under the same terms as R:

  "geoR is available as Free Software under the terms of the Free Software 
   Foundation's GNU General Public License."
   As free package, geoR is supplied without any warranty. 

  History:
  --------
  geoR has been developed over the last two years in the Dept of Maths 
  and Stats, Lancaster University - UK.
  
  During this period it has being used for teaching short courses and 
  for data analysis. Some scientific publications has been written, 
  using the resources provided by the package.
  
  The package has been available on the web and there is already a 
  reasonable number of users, considering the e-mails received about 
  usage/bugs/help.


  How does it differ from the current resources?
  ----------------------------------------------
  geoR differs in some aspects of the currently available tools for 
  geostatistical data analysis: the package sgeostats and the functions 
  included in the VR's spatial package.
  
  The main differences are the implementation of methods for likelihood 
  based estimation and Bayesian prediction of (transformed) Gaussian 
  random fields.


  The target audience:
  ---------------------
  This is always hard to predict. 
  Based on the feedback we've got so far we can say that the package 
  has been mainly used as:
  - teaching support tool for courses on geostatistics
  - data analysis tool for statisticians working on applied problems 
    as well as researchers in soil and environmental sciences, 
    biology, among others
  - support tool for simulation studies


  Extra documentation:
  ---------------------
  The package web page is: 
                www.maths.lancs.ac.uk/~ribeiro/geoR.html

  For a quick visualization of its functionality:
                www.maths.lancs.ac.uk/~ribeiro/geoRintro.html

  A pdf version of the latter is included in the package distribution as: 
                docs/geoRintro.pdf


  The distribution file and tests:
  --------------------------------
  The current distribution includes documentation for all the functions.
  It has passed "R CMD check" (R-devel, 1.3.0)


  Future developments:
  --------------------
  The file "docs/ToDo.geoR" has a list of features which should be 
  implemented in short/medium term.


  Related resources/work:
  -----------------------
  There is a joint work in progress with Ole Christensen (Aalborg 
  University, Denmark) to implement non-Gaussian spatial models.
  It's not yet clear whether this will results in a package on its own 
  or will be an expansion of the geoR functionality. 
  Ole has given a great contribution for geoR development, and we 
  therefore expect total compatibility of the resulting product 
  with geoR.

  Martin Schlather (Bayreuth University, Germany) is writing a package 
  for simulation of Gaussian random fields. We will be in touch with 
  Martin trying to improve the compatibility between the two packages.


                             Paulo J. Ribeiro Jr and Peter J. Diggle
                                               Lancaster, 25/05/2001

