Changes in 1.9-0

  * DESCRIPTION, NAMESPACE: import psychotools, new URL
  * soundquality (new data set and replication code): quality of multichannel
    reproduced sound
  * Code efficiency: replace all left-over occurrences of log(choose()) by
    lchoose() so group.test() and eba.order() handle larger data sets (spotted
    by Frank Papenmeier)

Changes in 1.8-0

  * schoolsubjects (new data set): preference for school subjects
  * circular (new features): exact p-values for small n, one- and two-sided
    tests, simulated p-values (prompted by Maarten Wijntjes and Cristina
    Spoiala)
  * kendall.u.Rd: original example from Kendall and Babington Smith (1940)
  * Documentation: add DOIs to references

Changes in 1.7-2

  * New functions: deviance.eba(), deviance.eba.order()
  * anova.eba: now works with stats::print.anova()
  * NAMESPACE: import graphics

Changes in 1.7-1

  * trineq (new function): check trinary inequality
  * inclusion.rule (new function): check inclusion rule (tree structure of
    aspect sets)
  * nobs.eba (new function): extract number of pairs
  * mbt.R: computes a bootstrap p-value; the ordering of the ranking data is
    now irrelevant
  * NAMESPACE: include OptiPt, trineq, inclusion.rule

Changes in 1.7-0

  * mbt.R (new function): Mallows-Bradley-Terry model
  * tartness (new data set): tartness rankings of salad dressings
  * boot.R: fix CRAN check warning about partial matching
  * OptiPt.Rd (references): add Luce (1959)
  * NAMESPACE: new file

Changes in 1.6-0

  * BUG FIX: use correct design matrix in thurstone.R
  * uscale (new function): utility scale extractor function for eba objects
  * cov.u: gains a norm argument for normalized utility scale values
  * EBA_fast.R, eba.order.R: objects of class eba and eba.order gain a
    coefficients component, so coef and confint methods are available; and
    they lose their obsolete se and ci95 components
  * OptiPt.Rd (examples): new plot, confint
  * heaviness.Rd (examples): new plot
  * drugrisk (examples): new plot
  * simulate.eba.Rd (examples): update

Changes in 1.5-7

  * BUG FIX: p-value in circular.R was incorrect (reported by
    Wolfgang Ellermeier)
  * drugrisk (new data set): perceived health risk of drugs
  * circular.R: remove diagonal before counting circular triads
  * DESCRIPTION: change depends on nlme to imports nlme

Changes in 1.5-6

  * simulate.eba (new function): simulate responses from EBA models
  * pcX: per default omits reference category
  * eba.order.R: eba.order(M1, M2=NULL, ...) accepts a 3d array
  * celebrities, heaviness, pork: more informative dimnames
  * linear2btl.Rd: add weighted least squares example

Changes in 1.5-5

  * strans.R (bug fix): replace viol.tab by violdf, better obj.names

Changes in 1.5-4

  * CITATION: add reference to BRMIC paper
  * strans.Rd: update value, examples

Changes in 1.5-3

  * strans.R: report the triples that violate transitivities, perform
    likelihood ratio tests of significance of WST violations
  * print.strans: update to accommodate new results from strans
  * eba.order.R: fix warning that occurred when taking square root of
    diagonal of Hessian; rather: sqrt of last element only
  * strans.Rd: update details, value, references
  * boot.Rd: delete unmatched right brace on line 32 that caused a CRAN
    check warning

Changes in 1.5-2

  * linear2btl.R (new function): transforms linear model coefficients to BTL
    parameter estimates
  * vcov.eba.order (new function): covariance matrix of eba.order model
    parameters
  * anova.eba: make it work for eba.order models
  * print.group.test, print.summary.eba, and print.summary.eba.order: delete
    the symbolic.cor=p>4 argument that caused a NOTE from CRAN check
  * pcX.Rd (see also): update

Changes in 1.5-1

  * balanced.pcdesign.R (new function): creates a (completely) balanced
    paired-comparison design
  * thurstone.R (new function): fits a Thurstone-Mosteller model (Case V)
    to paired-comparison data by maximum likelihood
  * anova.eba (new function): likelihood ratio test for eba models
  * vcov.eba (new function): covariance matrix of eba model parameters
  * zzMethods.R: add anova.eba.order
  * OptiPt.Rd: put anova and vcov
  * pcX.Rd: put reference to Critchlow & Fligner (1991)

Changes in 1.5-0

  * eba.order.R (new set of functions): fitting and testing EBA models
    that include a within-pair order effect
  * OptiPt: add names to the parameter estimates
  * print.strans: replace printCoefmat by print, now it prints
    "0" (not "NaN") when there are no SST violations
  * strans.Rd (value): update
  * heaviness.Rd (source): put original reference to the data

Changes in 1.4-3

  * BUG FIX: group.test now returns the correct log-likelihoods (the
    combinatorial constant is added to the models for the *pooled* data
    and not, as it was in 1.4-2, to the group model; this leaves the
    deviances unchanged, but the logLik's are now correct)
  * circular.R (new function): number of circular triads and indices of
    inconsistency
  * L.constrained (new function): constrained minimization
  * eba.boot.constrained (new function): bootstrapping using constrained
    minimization
  * OptiPt: add "constrained" argument for constrained minimization
    defaults to TRUE, parameters are constrained to be positive
  * boot.R: add "constrained" argument, see OptiPt
    add try() method in order to catch nlm errors
  * group.test: add "constrained" argument, see OptiPt
  * cov.u: add dimnames
  * weights data: rename to heaviness (weights is an existing function name)
  * plot.eba: add ... argument to plot
  * OptiPt.Rd (arguments, usage: method for summary.eba): update
  * boot.Rd (arguments): update
  * plot.eba.Rd (arguments, examples, usage: method): update
  * residuals.eba.Rd (usage, method): update
  * group.test.Rd (details, references): update

Changes in 1.4-2

  * kendall.u.R (new function): Kendall's u coefficient of agreement
  * eba (new function): wrapper for OptiPt
  * logLik.eba (new function)
  * weights (new data set): illustrate choice models with order effects
  * group.test: add combinatorial constant to logLik of all group models
  * Check strans for matrices: bug? No, returns NaN if there are no
    violations
  * OptiPt: add dimnames for fitted eba objects
  * celebrities data: add rownames
  * pork data: add dimnames
  * group.test.Rd (details): update
  * OptiPt.Rd (usage, arguments, details): update
  * strans.Rd (value, references): update
  * celebrities.Rd (description, source, examples): update
  * pork.Rd (description, source, examples): update
