CHANGES

Version 1.30.2

* Updated the documentation for some of the functions to concur with CRAN's requests and added @export tags for functions where it was missing.


Version 1.22

o Platform checking in MCmcmc updated, and checks for platform and
  package availability is now subsequent to printing of a BUGS program
  using argument code.only=TRUE

Version 1.21

o Bug in converting a MCmcmc object correctly to MethComp format if
  more than 2 methods were present. Now fixed.

Version 1.20

o VC.est.fixed updated with the possibility of having the MxI random
  effects correlated within items.

o A *SERIOUS* bug in VC.est.fixed and VC.est.random have been
  corrected: The estimated residual variances were shown in
  alphabetical order of the method names instead of in the same order
  as the method levels. Hence analyses by AltReg and BA.est prior to
  version 1.20 in datasets where method names are not in alphabetical
  order will be wrong.

Version 1.19

o New 'predict' method for Passing-Bablok regression (PBreg). It 
  follows the syntax of predict.lm (to an extent) and allows for
  calculation of confidence intervals.

o Updated plot method for PBreg - now shows correct confidence boundaries
  for the fit.  

o DA.reg updated so that the LoA is included too. Documentation
  accordingly updated.

Version 1.18

o Use of the alpha= argument crashed lines.MethComp (and all the
  functions that call this). Fixed.

o Checking for coda, R2WinBUGS and rjags improved to produce fewer
  weird error messages.

Version 1.17

o An argument mult=FALSE is added to BA.plot (and to plot.MethComp,
  lines.MethComp and points.Methcomp), which for Bland-Altman plots
  plots ratios on a log-scale instead of differences on a linear scale.

o The ridiculous functions min.Meth, max.Meth and median.Meth removed
  again. Only mean.Meth and sort.Meth that takes means / sorts
  measurements within item/replicate are preserved.

Version 1.16

o mean.Meth was not exported, it is now.

Version 1.15

o BA.plot is now in a new version essentially usng the facilities
  available from at MethComp-object derived from DA.reg.
  The old version is still available as AB.plot.

o JAGS is now recognized as a value for the program= argument to
  MCmcmc. Jags and rjags is required to be installed of course.

o More intuitive example in the Deming help page, thanks to Roland Fuss.

Version 1.14

o Further bug-fixes in VC.est.random getting the estimated variances
  right. (CTEk)

Version 1.13

o Bug-fixes in VC.est.random getting the estimated variances right. (CTEk)

Version 1.12

o Complete rewrite of BA.plot, the old still available, the new
  version is so far called AB.plot. AB.plot creates a MethComp object
  and uses plot.MethComp to make the Bland-Altman plot. The old
  workhorse BlandAltman will still be available for direct use.  This
  also involved changes in the output object from AltReg. (BxC)

o Amendment of plot.MethComp, to allow for increasing sd etc.

o Amendment of DA.reg to allow increasing sd, change in the returned
  object.

o New functions DA2y and y2DA to convert between linear relation
  between difference and average and linear relation between y1 and y2.

Version 1.11

o Added the dataset Ancona, collected at the MethComp course in Ancona
  on 28 September 2011.

o Changed the mode of the "random.raters" argument to logical
  throughout, preventing side effects of crap coding.

o DA.reg now returns the intercept and the slope of the differences as
  a function of the averages.

o A convenience function DA2y is added, converting the regression of
  the differences on the averages to the lines linking y1 and y2.

Version 1.10

o Removed warning from VC.est's call to lme wth a -1 in the model
  formula (which returns 0 d.f. for the fixed item parameters).

o BA.est named the 3rd column in the result as "sd", whereas all other
  functions returned the name of the column as "sd.pred". This caused a
  crash when using lines.MethComp, which assumes that "sd.pred" is the
  name of the third column. Now all called "sd.pred".

Version 1.9

o Incorporation of the effects of random.rates in the calculation of
  prediction limits by BA.est.

Version 1.8

o Claus Ekstrøm added the possibility of random raters to VC.est.
  In reality a new function is added, VC.est.random, the old is
  renamed to VC.est.fixed, and a wrapper VC.est is introduced.

o Printing of MCmcmc objects crashed because the inverse
  transformation was apllied also in cases where it was non-existent.

Version 1.7

o VarCorr is now explicitly called as the version the from nlme, so
  that it is not accidentally masked by the one for lme4 if that sould
  be loaded.

Version 1.6

o Printing of the transformation used (if any) is now working.

o Added Passing-Bablok regression, written by Michal Figurski, functions
  PBreg, with a print and a plot method for the resulting "PBreg" class.

Version 1.5

o Fixed the real bug which was in Meth. Now removes the columns with
  measurements by different methods when converting from a wide-form
  data frame to a Meth object.

Version 1.4

o Fixed a bug in DA.reg, preventing transformed analysis to be made if
  original wide data were carried over to the Meth object.

Version 1.3

o Fixed a few bugs in MethComp that prevented plotting of MethComp
  applied to MCMmcmc objects

o Enhanced DA.reg slightly to also show the estimated dependency of
  the residual sd. on the avearge.

Version 1.2

o Added Passing-Bablok regression, written by Michal Figursky, functions
  PBreg, with a print and a plotmethod for the resulting "PBreg" class.

Version 1.1.0

o Removed the possibility to explicitly give the working directory for the
  BUGS program. Due to problems in R2WinBUGS's openbugs function.

Version 1.0.1

o Improved output from Meth.

o Argument names have been made consistent between plot.MethComp and
  plot.MCmcmc, the common ones are now called 'points=', 'wh.cmp='.

Version 1.0.0

o Meth function, defining a Meth object have been rewritten, using a different
  syntax. Now allowing vectors to be combined into a Meth object, and arbitratry
  columns of a dataframe to be selected for the four essential columns of a
  Meth object.

In version 0.6.0

o Transform= argument added to BA.est, AltReg og MCmcmc

o MethComp-class, defined by a MethComp function added, handling printing
  and plotting of results from BA.est, AltReg and MCmcmc.

Version 0.5.2

o The way the two methods are ordered in BA.plot is reversed. Like in BA.est and
  VC.est, the FIRST method is now the reference; LoA are thus for M2-M1.

o summary.Meth() gives the median and range of measurements for each method too.

o The result from AltReg now have names for the dimnames to improve readability.

o A sample function for sampling the items from Meth object is added.

Version 0.5.1

o A Meth class has been defined, making sure that meth, item and repl
  all are factors. It has summary, print, plot, subset and summary methods.
  The summary method was formerly called tab.repl.

o A function has.repl() testing wheter a data frame has replicates or not.

o A DA.reg function added which estimates a linear relation between methods
  from the regression of the differences on the averages.

o plot.MethComp is now a proper S3 method for MethComp objects.
  The former functions plot.MethComp and plot.VarComp have been renamed
  plotMethComp and plotVarComp and are called by plot.MethComp via the
  what= argument.

o A function check.MethComp added, which gives convergence diagnostics
  for a MethComp object, and optionally calls the trace.MethComp and
  post.MethComp functions.

o summary methods for output from BA.est, AltReg og MethComp,
  which produces roughly same-style layout.

--------------------------------------------------------------------------------

Changes in 0.4.2

o Three new datasets from published papers added: VitCap, PEFR and
  Enzyme.


Changes in 0.4.1

o Argument random= in MethComp has been replaced by the arguments
  linked=, IxR, MxI.

o Aregument varMxI= added to MethComp allowing to constrain matrix
  effects to have identical variances across methods.

Changes in 0.4.0

o AltReg for fitting the general model iteratively is added.
  The estimation workhorse is now separated from BA.est to a special
  utility function VC.est, taht estimates in constant bias models.

o VC.est has been taken out from BA.est to form a separate workhorse
  also usable in the estimation iteration in AltReg. By default (also
  in BA.est) the variance of the matrix effect is now constant across
  methods.

Changes in 0.3.0

o New functions trace.MethComp and post.MethComp, giving the trace and
  posterior densities for the parameters separately for the four chains.

o It is now possible to use BRugs(OpenBugs) for the simulations, by
  specifying program="OpenBugs", "OB" or "Brugs" (case insensitive)
  in the call to MethComp. If nothing is specified MethComp will check
  if BRugs is installed and if so use it, otherwise attempt to use WinBUGS.

Changes in 0.2.2

o The BA.est function has been changed so that the LoA are based on
  p/m 2.00 instead of p/m 1.96.

Changes in 0.2.0

o The MethComp function revised and made somewhat modular. Default is
  now 4 chains, using different starting values for the variance components.

o The BA.est function is now used to generate sensible (and diverse) starting
  values for the variance components.

o A MethComp object is now of class mcmc.list, in order to allow use
  of the diagnostic tools from the coda package. Thus the posterior is now in
  a mcmc.list object (a list of matrices), transformed from the raw output
  mcmc.list object.

o R2WinBUGS and coda are not required any more, only suggested. Hence the
  BlandAltman plots etc. can be used without trouble loading the package even
  when R2WinBUGS and coda are not installed.

Changes in 0.1.19

o Typos in the documentation fixed (to some extent).

o plot.VarComp now takes range(grid) as the extent of the x-axes.

o Error in column naming in to.wide fixed.

Changes in 0.1.16

o BA.est has been expanded to handle different matrix effects (item by
  method) between methods if the number of methods exceeds 2. Thanks to
  advice from Peter Dalgaard.

o plot.meth now displays plots correctly. An argument to display the variable
  names inside the panels have been added.

o BlandAltman (and hence BA.plot) has been enhanced to accept axis
  specifications from limx, limy, eqax and ymax.

o BA.plot now does not give par() suggestions any more.

Changes in 0.1.13

o The default for the y-limit in the BlandAltman function now defaults to
  that the y-axis should have the same extent as the x-axis, symmetrically
  around 0, unless the differences or the limits of agreement are larger.