Changes in 1.11.13 (2012-02-25)

  o modifications to accommodate single-probe probesets without errors
  o sigma2.method default to "robust" in functions RPA.sigma2.update and rpa.fit
  o changed defaults in set.alpha function
  o added missing data imputation in rpa.fit

Changes in 1.11.12 (2011-12-13)

  o removed cind option from update.hyperparameters in RPA.online

Changes in 1.11.05 (2011-11-11)

  o online functions work properly

Changes in 1.9.19 (2011-10-29)

  o removed compiler and corrected the broken compiler function
    references

Changes in 1.9.17 (2011-06-27)

  o added bg correction in rpa.online
 
  o accurate and relatively fast variance hyperparameter update
    function added (see update.hyperparameters, update.s2)

Changes in 1.9.11 (2011-06-02)

  o speedups in hyperparameter estimation; alpha treated as scalar
  o removed affinity.method
  o added quantiles.online normalization method

BUG FIXES

  o alpha, beta updates fixed
  o rpa.fit did not return updated alpha, beta: now corrected.

Changes in 1.9.06 (2011-05-29)

  o rpa.fit-class: added alpha, beta prior parameters for the inverse
    Gamma conjugate prior for the probe-specific variances

  o modified RPA.iteration, RPA.update.sigma2 so as to directly
  utilize priors everywhere. 

  o added set.alpha, set.beta and update.hyperparameters,
  update.alpha, update.beta in internal functions

  o added toydata generator function sample.probeset

  o rpa.plot: added comparison of toydata and fitted data by adding
  the toydata.object argument

Changes in 1.9.03 (2011-05-14)

  o corrected AffyCompII result note

Changes in 1.9.03 (2011-04-27)

  o polished plot functions in rpa.plot and plot-methods

Changes in 1.9.02 (2011-04-21)

  o added compiler to betahat in RPA.sigma2.update.R to speed up

  o added NA/NaN check to RPA.iteration after a defected affybatch
    with NA values was found to cause crash

Changes in 1.7.34 (2011-03-14)

  o added note on AffyComp results in the Vignette

Changes in 1.7.32 (2011-03-12)

SIGNIFICANT USER-VISIBLE CHANGES 

  o added estimation of probe affinities in rpa and RPA.pointestimate
  functions. Now returning expression values in the original data
  domain, i.e. in 'absolute' expression levels. Provides options for
  affinity estimation through affinity.method field in certain functions.
  o added 'affinity' field in rpa-class and rpa.list-class
  o added estimate.affinities function
  o removed exclude reference array option
  o added rpa.fit class (S3) to provide storage for an individual model
  o rpa.fit function now fits the model (including affinities) for
    given data. This is a generic function to calculate an individual
    model instance for any data following the distributional
    assumptions of the model. See tests/rpa.fit.test.R for an example. 

Changes in 1.7.3 (2011-01-20)

BUG FIXES

  o in 'rpa' wrapper function the exclude.reference.array option was
    interpreted in the opposite way than intended. Now the
    reference.control.array = TRUE excludes the reference array from
    output. Default is FALSE, i.e. reference array is included.

Changes in 1.7.2 (2011-01-14)

  o updated citation
  o moved license from GPL>=2 into FreeBSD since that is less
  restrictive open license. OK since all copyrights of the GPL>=2
  version are with the package author.

Changes in 1.5.2 (2010-05-19)

USER-VISIBLE CHANGES
  o the vignette has been improved
  o added firstlib.R function to print copyright notice during loading
  o added a copy of the GNU GPL 2 licence
  o added citation information, see citation("RPA")

Changes in 1.3.6 (2010-03-28)

BUG FIXES
  o added documentation on 'get.probe.noise.estimates' function

Changes in 1.3.5 (2010-03-18)

USER-VISIBLE CHANGES

  o in the 'rpa' preprocessing wrapper, include the reference array
  (cind) to the differential gene expression matrix
  output. Optionally, the reference array can be excluded. See the
  'include.reference.array' parameter of function 'rpa'. Note that all
  differential expression values of the reference array are 0 since
  the reference is compared against itself.

  o added function get.probe.noise.estimates. Outputs probe-level
  noise estimates with normalization options.

BUG FIXES

  o added destructive=TRUE into background correction. Should help with
  large data.


Changes in 1.3.2 (2010-03-02)

SIGNIFICANT USER-VISIBLE CHANGES

  o changed default sigma2.method into 'robust'

  o control array default set to cind = 1 (previosly sampled at
  random if not specified)

  o initialize.priors replaced by set.priors function


USER-VISIBLE CHANGES

  o removed sigma2.guess from user-defined parameters in RPA.iteration

  o RPA.pointestimate: added regularization parameters alpha, beta


NEW FEATURES

  o support for alternative CDF files

  o rpa function: a wrapper for preprocessing gene expression data
  with the RPA model

  o rpa.plot function: visualize probe-level data and results for a
  given probeset

  o background correction to RPA.preprocess

  o option to define background correction and normalization
  parameters

  o regularization for noise term estimation (sigma2)

  o sigma2.method 'mean' and 'robust' added


BUG FIXES

  o rpa class: added elements 'data', 'cdf' and 'abatch'

  o RPA.preprocess: background correction added, and log2
  transformation now done after normalization

  o added 'maxloop' parameter to RPA.iteration function call 

  o moved the contents of R/zzz.R into AllGenerics.R, AllClasses.R,
  *-methods.R, *-accessors.R

  o code optimized for speed