Dear Emacs, please make this -*-Text-*- mode!
     
     CHANGES IN  Renext package VERSION 3.0-0	

CHANGES

   o Only minor changes in doc/guide. The major number stands for the
   completion 2.x efforts.

   o Remind that 2.1-10 on CRAN was a patched version solving problems
   revealed by CRAN checks

     CHANGES IN  Renext package VERSION 2.4-0	

CHANGES

   o Several functions involving POSIXct objects have been changed in
   the aim of removing concatenation of POSIXct with the 'c' method,
   which drops the timezone and leads to strange timezone in the
   remote past.

   o In the 'OT2MAX' function, the code for the preparation of the
   output was unclear for the case when 'dataFrames' is FALSE. 

   o Changed the doc of the 'parDeriv' function.

   o in 'fGEV.MAX' a suitable constant is added to the final logLik
   in order make this value comparable to the one computed by the 
   classical method for maxima or r largest (evd, ismev).

BUG FIX

    o In 'plot.renouv', the 'pct.conf' argument was not duely passed
    to 'lines.Renouv'. Fixed.

    o In 'OT2MAX' function the determination of the end was wrong
    because it was loosing a full year at the end of data.

     CHANGES IN  Renext package VERSION 2.3-0

CHANGES

   o The content of the 'data' directory was renewed to save the
   examples as '.rda' files instead of the previous '.R' dump files.
   The 'POSIXct' objects in data are now with "GMT" time zone. The .Rd
   files were also modified to change some POSIXct extractions and to
   remove the 'AsIs' class indication for chr variables ('comment').
   
   o 'rRenouv' and its help is removed. 
	
   o 'gpd2lomax', 'gpd2maxlo', 'lomax2gpd', 'maxlo2gpd' are no longer
   exported and their .Rd files have been removed.

   o Code and doc of 'LRExp.test' function changed to add the "maxlo"
   alternative. Code for 'anova.Renouv' now calls 'LRExp.test' when
   needed, to avoid code duplication.
 
   o 'readXML' and its documentation: added the possibility of
   specifying the time zone. Doc: browse 'index.xml' in the examples.

   o Typos and misspellings in the doc.

   o Help page for 'NBLevy' (outdated reference and so on).

    
BUG FIX

   o In 'Renouv' the correction of initial values for the "GPD"
   distribution with initial shape < 0 was not done.

   o In 'lines.Renouv', the DEBUG parameter was TRUE generating
   verbose noise.

    CHANGES IN Renext package VERSION 2.2-3

CHANGES

   o Added derivation w.r.t. parameters of the Lomax and Maxlo
   distributions in 'parDeriv', and changed the 'examples' section.

   o In the 'predict' method for 'Renouv' objects, the probability 
   column will no longer be found by default. The previous computation
   can be retrieved by using the non-default formal value prob = TRUE.
   
   o Some typos/errors in the documentation have been fixed.

   o Replaced 'require' by 'requireNamepace' for suggested package
   "XML".

BUG FIX

   o For the 'Lomax' and 'Maxlo' distributions 'plomax' and 'pmaxlo',
   the parameter 'lower.tail' was misleading, since they did not work
   as usual. Yet the default was for the distribution function as 
   usual so this problem had no impact on ML estimation. 


     CHANGES IN Renext package VERSION 2.2-2

CHANGES

   o 'predict.Renouv' now has its argument 'cov.rate' working.
   Turn to TRUE for default?

   o added some 'eval' needed in the legend construction
   'lines.Renouv'.

   o Corrected inefficient data control in 'makeOTSdata'.
     

     CHANGES IN Renext package VERSION 2.2-1

BUG FIX

    o Problems with symbol numbers and names in 'lines.Renouv'.

             CHANGES IN Renext package VERSION 2.2-0

BUG FIX

    o With the "GPD" distribution 'Renouv' did not return the logLik
    as expected.

CHANGES
	     
    o The package now imports from the 'splines' package since
    'splineDesign' is required to approximate the quantile functions
    required in some tests.

    o Technical data is added in a 'sysdata.rda' file of the R
    directory.
		 
    o In 'Renouv', the returned data list named 'history.MAX' and
    'history.OTS' have two new slots called 'dataNames' and
    'blockNames'.  The are extracted from the object's data frames
    when 'x' has class "Rendata" or are extracted from the names used
    in the list provided by user in the 'MAX.data' or 'OTS.data'
    arguments when 'x' is numeric. These indications are used in the
    legend built for plots built with the 'plot' or 'lines' methods
    of the "Renouv" class.

    o The the formal arguments of 'plot.Renouv' and 'lines.Renouv'
    have been modified to allow the user to better manage the block
    data (MAX or OTS).  Some blocks can be shown or not thanks to
    extended possibilities of the 'show' argument. The new
    'byBlockStyle' argument can also be used to decide if the blocks
    of a same type (MAX, OTS) should be plot with the same style 
    and to adapt the legend in correspondence. 	  
    
    o Many changes have been done in the 'RenextGuide' document shipped
    as a pdf file with the package. 

NEW FEATURES
    
    o Several functions have been added to perform tests of
    exponentiality or Gumbelity: 'CV2.test, 'Jackson.test',
    'LRExp.test', 'LRGumbel.test' and other related functions,
    e.g. the 'pGreenwood1' function. The related 'anova' method
    was also implemented.


     CHANGES IN Renext package VERSION 2.1-9

BUG FIX

    o The new 'OT2MAX' did not work without missing information.
 

     CHANGES IN Renext package VERSION 2.1-8

BUG FIX

    o The 'SandT' function did not work correctly for objects with
    class "Renouv".

NEW FEATURES 

    o The 'OT2MAX' now returns some summary information about gaps and
    can plot the mean gap rate.


       CHANGES IN Renext package VERSION 2.1-7

CHANGES 

    o 'Hpoints' is now duely exported.

    o  Corrections in 'fmaxlo' and 'fGPD' docs concerning the constraint.

BUG FIX

    o The 'flomax' and 'fmaxlo' functions could not work with cov =
    FALSE, due to a bad variable name. Fixed.


        CHANGES IN Renext package VERSION 2.1-6

NEW FEATURES 

    o It is now possible to use the 'RLplot' method on an object of
    class "Rendata". The plotting positions come from 'SandT' and no
    fit is required.

CHANGES 

    o 'SandT' now optionally uses the old plotting positions with
    naive = TRUE.  	
    
    o 'lines' 'PPplot' and 'QQplot' methods now use the plotting
    positions as described in 'Renext Computing Details'.

    o The maximal number of historical blocks to be coloured
    individually in 'plot.Renouv' or 'RLplot' can be reduced.

BUG FIX

    o In the function 'SandT', the removal of ex-aequos among 
    the tresholds was wrong (and finally useless). Fixed.
    
    o When fitting the "GPD" special distribution with historical
    data, the check that all HD falls in the support was missing. 

         CHANGES IN Renext package VERSION 2.1-5

CHANGES 

    o 'fmaxlo' now uses a lower bound on the shape parameter, since
    the unconstrained ML estimate is always at the boundary.

    o In 'flomax' a better lower bound for the ML estimate of the
    scale parameter is now used.

    o 'fmaxlo' and 'flomax' now use an explicit formula for the
    inverse of the expected information matrix. Both functions now
    have a 'cov' formal. When FALSE, the computations boil down to a
    maximisation of the profile log-likelihood, and the covariance is
    not computed. This aims to have a fast evaluation e.g. in order to
    sample a likelihood ratio statistic.
 
    o 'fGPD' has been refactored, and now merely turn to the
    estimation of an exponential distribution when the CV is close to
    1.0.

BUG FIX

    o In the function 'SandT', the value of the survival was incorrect
    because it used the rate of events for the first inter-threshold range
    only, and not the whole rate.

          CHANGES IN Renext package VERSION 2.1-4

CHANGES 

    o Added a new distribution named "GPD" that can be used in
    'Renouv'. This distribution has functions returning NaN when the
    parameter vector falls outside of its validity domain (scale > 0),
    and therefore unconstrained optim can be used for ML estimation. 

    o In 'fGEV.MAX', a scaling is added and a 'try' to overcome the
    problem of non-invertible hessian. A new threshold choice is
    proposed based on the spacings.
       
BUG FIX

    o In the function 'fmaxlo', the lower bound chosen for ML
    estimation was too small for a numerical evaluation of logLik and
    its derivative, leading to a (wrong) negative derivative, and to a
    warning.

           CHANGES IN Renext package VERSION 2.1-3

CHANGES 

    o The 'envir' formal of the 'RLlegend.*' functions is removed.
    for simplicity. A new function 'RLlegend.print', non exported, is
    added for debugging legend problems.

    o A 'label' formal argument is added to 'Renouv' and is by default
    set to "". This character string appears in the legend as a prefix
    to the graphical objects description ("quant", "conf", ...) and
    where a mysterious "res" was shown before. When the call to the 
    'plot' method within 'Renouv' leads to an error, the result
    is returned with a simple warning.
       
BUG FIX

    o In the function 'makeOTSdata', some problem could occur
    when MAX.r == 0. This created strange problems in 'SandT'.

           CHANGES IN Renext package VERSION 2.1-2

CHANGES 

    o New functions dedicated to block maxima in relation with
    POT/Marked process. These are 'gGEV.MAX' and utility functions
    'Ren2*', '*2Ren', 'parini.MAX' and 'parini.OTS', ...
    
    o New function 'rRendata' to simulate "Rendata" objects. This is
    very useful for testing estimation of plotting functions.    

    o New method 'spacings' computing the spacings.
	
    o New function 'fGPD' for ML estimation of a GPD from a sample. Works
    by calling either 'flomax' or 'fmaxlo'.

    o New function 'parDeriv' computing the derivatives w.r.t parameters
    of the density and survival of some distributions, at the time,
    only "weibull" and "gpd".

    o In 'Renouv' it is now possible to give 'x = NULL' in which case only 
    historical data are used. At the time, this is done by calling the 
    'Renouv4' function. The 'parini.MAX' and 'parini.OTS' functions
    perform the key step of parameter initialisation.
    
    o New function 'OT2MAX' to create block maxima from marked
    process data.
    
    o New function 'SandT' computing the survival and the return periods
    using an adaptation of Hisch and Stedinger's method. This is 
    useful for plotting positions. 

    o New methods 'QQplot' and 'PPlot' for the class "Renouv". Some
    work is still needed to deal with historical data.


           CHANGES IN Renext package VERSION 2.1-1

BUG FIX

    o 'print.summary.Renouv' produced an error when some OTS blocks
    had length 1. Fixed.

CHANGES

    o Following an advice given by B. Ripley, an environment object is
    created in the package in order to share variables between the
    package functions without turning these into global
    variables. Thus the legend management functions 'RLlegend.ini',
    'RLlegend.show' and the 'plot' and 'lines' method for class
    "Renouv" no longer require defining an environment and passing it
    as an argument. See the example of 'RLlegend'.
 
	         CHANGES IN Renext package VERSION 2.1-0

CHANGES

    o Global variable 'RLlegend' renamed to '.RLlegend' and added a
    "Caution" section in the help page of 'plot' and 'lines' methods  
    for the S3 class "Renouv".
	
	         CHANGES IN Renext package VERSION 2.0-15

CHANGES

    o Changed the 'Depends' Imports' and 'Suggests' lists of packages.

    o 'Renouv' replaced by 'Renouv3' (renamed).
    
    o Many slight changes in the manual, some in the 'Renext Guide'.


BUG FIX

    o 'df' misleadingly returned the number of obs in 'RenouvNoEst'.


	         CHANGES IN Renext package VERSION 2.0-14

CHANGES

    o Added an 'nobs' field in output of 'Renouv3' (for AIC or BIC). 
    
    o The 'logLik' method for class "Renouv" now uses 'df' and 'nobs'
    and no longer uses 'p.y' and 'nb.OT', but duly computed values
    which can be used with historical data. The documentation of the
    'logLik' method was changed consequently.


BUG FIX

    o The the number of obs was 'df' in the output of 'Renouv3'.


	         CHANGES IN Renext package VERSION 2.0-13

CHANGES

    o The functions 'roundPred' and 'plot.Renouv' can now work
    with prediction tables containing NA. This can occur if the 
    standard deviations of estimates are not available, e.g. 
    if the shape parameter of the GPD is < -0.5.
    
    o In 'Renouv3.R', some precautions have been added using 'try' in
    order to manage situations when the hessian is non-invertible or
    has negative eigenvalues. For compatibility reasons, the
    covariance and standard deviations objects are then filled with NA
    rather than set to NULL.


BUG FIX

    o When using the plot method with an object with S3 class
    "RenData" an error could happen with MAXdata containing NA
    dates. Fixed.


	         CHANGES IN Renext package VERSION 2.0-12


CHANGES

    o In the package DESCRIPTION, 'LazyData' was turned to 'yes'.
    Thus, the datasets 'Garonne', 'Brest', ... are now available
    without requiring "data" as in "data(Garonne)".

    o The 'fML' and 'Renouv3' functions now return the maximised
    log-likelihood as a 'logLik' list element. Bounds for the
    parameters are also attached to the distributions, but not used
    yet in the optimisation. The problem is that NaN are not supported
    in constrained optimisation methods but such values are likely to
    occur when maximising a gpd log-lik.

    o The probability functions for the "Lomax", "Maxlo" and "SLTW"
    distributions no longer produce an error when the parameters do
    not fill the sign constraints. Instead, a (vector of) NaN is
    returned. This allows the use of an unconstrained optimisation
    method such as "BFGS".

    o The functions 'flomax' and 'fmaxlo' dedicated to the MLE for the
    "Lomax" and the "Maxlo" distributions have been improved.  Bounds
    have been derived to localise the ML estimate, and a scaling is
    applied to the data used during the optimisation.  These two tasks
    remain to be done for the case where the shape parameter is known
    'flomax1' and 'fmaxlo1'.

    o Many small changes were made in the document 'RenextGuide'.
    These change are now described in a special 'ChangeLog' file in
    the 'doc' directory.

    o Slight corrections in .Rd files (e.g., 'RLlegend'). 

	      
NEW FEATURES

    o The "Renouv" S3 class now has methods 'logLik' 'nobs' and hence
    also 'AIC' and 'BIC'. However, these methods do not yet provide
    results when historical data are provided.  Some work is still
    needed to design the constants and to define 'nobs' for fits which
    use historical data.

BUGS FIX

    o In 'Renouv3', the arguments 'start.par.y' and 'fixed.par.y'
    are now intended to be named  vectors rather than lists so passing
    then as lists produced an error. Fixed.
      
    o The "maxlo" distribution could produce an error when used with
    historical data, because the first ML estimate of the scale 'beta'
    (est. without historical data) could be smaller than one
    historical data. Fixed.

    o In 'RenouvNoEst', an error occurred when "distname.y" was not a
    special distribution, because the parameter names were not duly
    retrieved from the 'estimate' arg. Fixed.
 

        	CHANGES IN Renext package VERSION 2.0-11

CHANGES

    o Unused hidden file (.BAKRbuildignore) removed.

    o Coerce a 'skip' argument to integer (read as character from XML
    file).

    o Changed the Maintainer address, now the personal address of Lise
    Bardet.

 	         CHANGES IN Renext package VERSION 2.0-3


NOTE

    o The package still exports two undocumented functions named
    'Renouv2' and 'Renouv3'. These functions have the same args as
    'Renouv' with the same meaning, but are refactored versions with
    some new features.  The function"Renouv3" is the one intended to
    replace "Renouv" in the next CRAN version.

NEW FEATURES

    o Method 'vcov' for the S3 class "Renouv" to extract the
    covariance matrix of estimates.
     
CHANGES 
	
    o The 'RLlegend' function no has an explicit argument 'bty' to
    overload the default value of 'legend'. Legends produced with
    'RLlegend' no longer show a box by default.

    o Changes in the "RenextGuide" document. They concern the
    likelihood for OTS data (subsection 3.4.2 "Likelihood", therein a
    remark about concentrating the likelihood with historical data and
    coefficient of variations (subsections B.2.2 "Some indicators" and
    B.3.2. "Generalised Pareto Distribution/ Estimation and
    inference"). This document does not tell much about the new
    features for versions > 2.0-0.
       
BUGS FIX

    o In 'plot.Renouv', the argument 'pct.conf' was not passed to
    'lines.Renouv' as it must be.



 	         CHANGES IN Renext package VERSION 2.0-2

NEW FEATURES

    o New functions 'RLlegend.ini' and 'RLlegend.show' to fill legends
    for return levels plot built with two or more 'plot' and 'lines'
    steps. The legend is built in a semi-automatic fashion by
    collecting information given in the 'par' argument: line types,
    colors, ...

    o New temporary function 'Renouv3'. This follows the re-factored
    'Renouv2', but the likelihood is now concentrated when historical
    data are given. More precisely, the event rate parameter 'lambda'
    is concentrated out of the likelihood using an appealing formula
    
    'lambda.hat' = 'number of events'/'discounted duration'
    
    where the duration of a historical block is discounted by 
    a survival value. 
     
          
CHANGES 
	
    o Reference manual: small corrections. The line 'library(Renext)'
    is removed from the examples.

    o The 'plot.Renouv' argument list was changed to be in accordance
    with that of 'lines .Renouv'.


BUGS FIX

    o In the 'Renouv2' function.



  	         CHANGES IN Renext package VERSION 2.0-1

NEW FEATURES

    o 'RenouvNoEst' function to create a new object of class "Renouv"
    without estimation.
    
    o 'lines.Renouv' function/method to add graphical elements from a
    "Renouv" object to an existing return level plot a used in Renext.

    o 'RLpar' function to change the value of the graphical parameters.

    o New ML estimation functions 'fgamma', 'flomax' and 'fmaxlo'
    (experimental and undocumented).

    o New function 'fgpd1' for the estimation of a GPD with given
     shape 'xi' (and location 'mu') but with unknown scale 'sigma'.


CHANGES 

    o Re-factoring of the 'Renouv' function in order to reach a better
    code re-use. The result is in a 'Renouv2' function still on
    development.


BUG (NOT FIXED YET)

    o In the 'RenextGuide' document the log-likelihood given for OTS
    historical blocks contains an error.

    o In the 'plot.Renouv' function/method, the pct.conf may not work
    as expected/wanted in this version. This is because more
    flexibility is now needed in the selection of the graphical
    elements that must be displayed in order to have a good interface
    for the 'lines.Renouv' function/method.
    
    o In the 'plot.Renouv' function/method, the adding a text create
    an error due to the use of an unbound variable 'ry'.


     	         CHANGES IN Renext package VERSION 2.0-0

CHANGES

    o This 'Changelog' cleaned.	


                CHANGES IN Renext package VERSION 1.5-1

CHANGES

    o Typos in the 'RenextGuide' document.

 
		CHANGES IN Renext package VERSION 1.5-0

CHANGES

    o The 'fit.Renouv' function is renamed 'Renouv' as would expect
    most R users.

    o The shifted Pareto 'spareto' distribution and its functions are
    renamed as "Lomax" distribution and 'xlomax' with x in "d", "p",
    "r", "h", "H".

    o Many slight changes in the vignette most of them striving to
    improve the English.

    o The vignette code (.Rnw) and subsequent material needed is no
    longer shipped with the package for the sake of its size and the
    resources needed in the build step.
    
    o The warning message produced when hessian is not definite
    negative is strengthened.


BUG FIXES

    o In the 'plot.Renouv' function, when historical data are shown on
    the plot it can be the case that their number exceeds the
    predicted number. This was a problem for the determination of the
    plotting position. Now the predicted number is modified in order
    to be at least the observed number.


		CHANGES IN Renext package VERSION 1.4-1

CHANGES

    o The 'fRenouv' function is REMOVED, and 'fit.Renouv' should now
    be used.

    o Many minor corrections (essentially spelling) were made in the
    documentation.

    o The documentation of the 'fit.Renouv' function now lists the
    'special distributions' and their parameters.

    o The 'print.summary.Renouv' function now informs the user about
    the inference method used (delta method or exact chi square). It
    also displays relevant information about historical blocks used in
    the fit, if any.
  
    o The 'ret.lev' table returned by 'fit.Renouv' no longer has
    rownames indicating the return period. The reason is that this
    table (a data.frame object) can not by default have ties in its
    rownames, and formating general (not 'pretty') return periods is
    difficult.

    o Changes in the vignette.  NOT FULLY CHECKED YET (spelling!)
    Some details have been added about the "log-exponential" and the
    "square-exponential" distributions.


BUG FIXES

    o The 'predict.Renouv' gave an error when the object 'x' had a
    transformation 'trans.y'. Fixed.

    o The 'fit.Renouv' function returned in some cases an error
    message in place of a table in its 'ret.lev' element.  This was
    due to 'rownames'. Fixed.



		CHANGES IN Renext package VERSION 1.4-0

CHANGES

    o The 'plot.Renouv' function was changed. The 'problim' and 'Tlim'
    arg now work as expected. The rule to set the x axis is quite
    complex to describe, but yet rather natural to use.

    o The vignette was modified mainly in the chapter devoted to the
    'fit.Renouv'. The two types of historical data 'MAX' and 'OTS' are
    now described with their likelihood. There is no longer reference
    to the old 'fRenouv' function, and all code chunks now refer to
    'fit.Renouv'. The methods for 'Rendata' and 'Renouv' classes are
    shortly described.


BUG FIXES

    o In 'fit.Renouv', a problem occurred when only one MAX block was
    given without being enclosed in a list().

    o In 'plot.Renouv' it was implicitly assumed that 'L.95' and
    'U.95' were present in the table which caused an error in the
    determination of the y axis.


		CHANGES IN Renext package VERSION 1.3-1

NEW FEATURES 		
			
    o 'fit.Renouv' now internally uses the function 'plot.Renouv' when
    its 'plot' formal is set to TRUE

    		   
CHANGES 

    o Packages 'evd', 'numDeriv' are now required 'Depends'.


BUG FIXES

    o In 'fit.Renouv', a bug affected 'historical data' arguments. It
    caused problems with "gpd" distribution with historical data.


		CHANGES IN Renext package VERSION 1.3-0

NEW FEATURES 		
			
    o 'fit.Renouv' now internally uses the function 'plot.Renouv' when
    its 'plot' formal is set to TRUE

    		   
CHANGES 

    o Packages 'evd', 'numDeriv' are now required 'Depends'.

    o In 'makeOTSdata' and 'makeOTSdata' giving a NULL value to the
    argument data allows to bypass the use of the historical data
    possibly present in the x object. A similar behaviour is know
    retained in the the 'fit.Renouv' function.

    o In 'plot.Renouv', the arguments have been changed. 'pct.conf',
    and arguments concerning marks suffixed as '.mark'. By default, no
    mark is shown. The arguments 'main' and 'ylim' have been removed
    but can still be used through the dots mechanism. The new
    arguments 'problim' and 'Tlim' allow the user to set the axis
    limits. When historical data are present in the object, they are
    now plotted with filled symbols, the filling color indicates the
    block.

	
BUG FIXES

    o In 'fit.Renouv' a problem occurred with MAX data when one block
    was given as a numeric vector (rather than a list).  A numeric
    problem occurred in with 'OTS' data with no threshold exceedances
    due to the use of r*log(1-F) with r = 0 and F very close to 1.
    


		CHANGES IN Renext package VERSION 1.2-0

NEW FEATURES 		
			
    o New function 'fit.Renouv'. This is mainly a cleaned version of
    'fRenouv' (which will remains here for a couple of sub-versions).
    The historical data are now clearly identified as of two types
    'MAX' and 'OTS', both using one or several periods called
    blocks. MAX data (or r-max) are as before with new formals; OTS
    data use blocks with all values over a known threshold over the
    main sample threshold.  The likelihood is then slightly different
    than that of the MAX-data case.
    		   
    o New S3 class 'Renouv' with methods: 'summary', 'coefficients',
    'predict' and 'plot'. The method 'predict' computes return levels
    and confidence bounds (when possible), while 'plot' produces a
    return level plot with the main sample and also historical data
    when present.
    
    o New method for the class 'Rendata': 'summary'. The 'plot' method
    was slightly changed to show historical data when this is
    possible.

    o New function 'OTjitter'. Works like 'jitter' but avoids
    jiterrized data falling under the threshold. This is used in
    'fit.Renouv' to avoid the warning concerning ties in the
    Kolmogorov-Smirnov test. In practice, there seem to very very
    little difference in the stat and p-value when jitterizing is
    used.


CHANGES 

    o Vignette. Correction of an error (p. 32) about the return-level.
    Modification of the parts concerning historical data.


		CHANGES IN Renext package VERSION 1.1-0

CHANGES 		
			
    o Function 'qmixexp2': quadratic interpolation during the first
    iterations replaced by a simpler linear interpolation, i.e. a
    standard Newton-Raphson method, easier to analyse. The method has
    been secured by placing the iterate below the true unknown root on
    the basis of concavity, by inflating the derivative if needed.
    Scale the data through a multiplication of these by the smallest
    rate names 'rate1', which should improve numerical accuracy.
    
    o Vignette. Correction of bugs in the index and around code
    chunks. The were due to end of lines chars that were moved in the
    .Rnw source file.

    o Vignette. Correction of several misspells or typos. Added a
    figure showing GPD densities for several values of the shape
    parameter 'xi'.


