-------------------- CDM NEWS --------------------------------

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

CDM package

A. Robitzsch, T. Kiefer, A. C. George, & A. Uenlue

mailto:
a.robitzsch /-\ bifie.at

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++


--------------------------------------------------------------
CDM 3.3 (2014-05-25)   --  CRAN CDM 3.3-24
--------------------------------------------------------------

x included structured latent class analysis for polytomous data ('slca')

* included/modified datasets: ---
* included/modified examples: slca (1,2,3,4,5)

--------------------------------------------------------------
CDM 3.2 (2014-04-23)   --  CRAN CDM 3.2-6
--------------------------------------------------------------

x included an argument save.devmin in 'gdina' for more parsimonious
  savings of output values to avoid working memory overflow for 
  large datasets
x included effect size of differential item functioning in
  gdina model (included in 'gdina.dif')
x fixed a bug in itemfit.sx2

* included/modified datasets: ---
* included/modified examples: gdina.dif (1)

--------------------------------------------------------------
CDM 3.1 (2014-03-21)   --  CRAN CDM 3.1-14
--------------------------------------------------------------

x fixed a bug when printing objects of class 'din'
x allowed vector input of item response pattern in 'plot.din'
x disabled argument log.file in 'summary.din'
x included random initial parameter values in 'din' and 'gdina'
x changed some settings in plot.din

* included/modified datasets: ---
* included/modified examples: din (1), gdina (1), plot.din

--------------------------------------------------------------
CDM 3.0 (2014-02-26)   --  CRAN CDM 3.0-29
--------------------------------------------------------------

x changed output in summary.gdina a bit (fixed a bug for
  the output of skill probabilities)
x included the argument 'avoid.zeroprobs' in gdina function
   to avoid zero probabilities of skill classes and inversion
   of singular matrices   
x fixed problems in itemfit.sx2 for empty cells (thanks to
   Balal Izanloo)
x included a Wald test for the GDINA model (gdina.wald)
x included assessment of item-wise differential item functioning 
   in the GDINA model (gdina.dif)
x removed minor bug in summary.gdina
x fixed a bug with singular design matrix in skill space reduction    
x included calculation of polychoric correlation in summary.gdina
   and an additional function skill.polychor
x removed parallel plot form plot.din
x fixed a (crucial) bug for in the multidimensional GDMs with item slope
  parameter estimation
x included function 'mcdina' for estimating the multiple group multiple
  choice DINA (MCDINA) model with polytomous attributes

* included/modified datasets: data.cdm01, data.cdm02
* included/modified examples: data.ecpe, gdina.wald (1), 
   gdina.dif (1), din (7), mcdina (1,2,3)

--------------------------------------------------------------
CDM 2.7 (2014-01-30)   --  CRAN CDM 2.7-7
--------------------------------------------------------------

x included cognitive diagnostic indices based on Kullback-Leibler
   information ('cdi.kli')
x changed NAMESPACE and DESCRIPTION files due to a change in Rcpp
x included non-exported Rcpp files for calculating individual 
  likelihood and posterior distribution for polytomous items and
  multiple groups

* included/modified datasets: ---
* included/modified examples: cdi.kli (1)

--------------------------------------------------------------
CDM 2.6 (2014-01-19)   --  CRAN CDM 2.6-13
--------------------------------------------------------------

x fixed a bug in polytomous GDINA models with only one skill
x included value 'zeroprob.skillclasses' in skillspace.hierarchy
x included a more efficient skill space reduction method in
  gdina: using the argument 'reduced.skillspace.method',
  the new implemented method is now the default
x included data transformation function 'sequential.items' for
  constructing dichtomous pseudo items from polytomous items

* included/modified datasets: data.sda6, data.timss03.G8.su,
  data.timss07.G4.lee
* included/modified examples: gdm (10)

--------------------------------------------------------------
CDM 2.5 (2014-01-16)   --  CRAN CDM 2.5-48
--------------------------------------------------------------

x fixed a bug when inverting a singular matrix in standard error
  calculation in gdina function (thanks to Hsin-Hao Chen)
x included conditional item probabilities in gdina output
x included gdina plot function (thanks to a discussion with 
  Hamdollah Ravand)
x fixed a bug in summary.dina if data without column names
  is provided (thanks to Hu Jinxiang)
x removed a bug in gdina function which occured sometimes
  due to division by zero (thanks to Hu Jinxiang)
x fixed a bug in cdm.est.class.accuracy for GDINA models
  (thanks to Hamdollah Ravand)
x fixed some instable numerical behavior for higher order
  GDINA models (thanks to Ed Merkle)
x fixed a bug in calculating number of estimated parameters for
  'din' function in case of skill classes with zero probability
  (thanks to Juergen Gross)
x added some explanations in 'modelfit.cor'
x included S-X2 item fit statistic ('itemfit.sx2') accompanied
  with a summary and a plot method
x included marginal skill classification reliability in
  'cdm.est.class.accuracy' function
x implemented significance tests in modelfit.cor and modelfit.cor.din
x included jackknife estimation of standard
  errors and bias corrected estimators in modelfit.cor.din
x speeded function 'modelfit.cor' by moving some parts of the code
  to Rcpp
x included warnings in case of nonconvergence in gdina function
  (thanks to a dataset provided by Hamdollah Ravand)
x slight change of algorithm for skill space reduction which
  seems to improve convergence behavior
x included a function for skill space approximation
x included a function for calculation reduced skill spaces
  if hierarchies among skills exist ('skillspace.hierarchy')
x included generalized distance discriminating method and a function
  for computing ideal response patterns

* included/modified datasets: data.hr, data.melab, data.jang, 
  data.dtmr
* included/modified examples: gdm (4,5,6), gdina (8,9), sim.din (3), 
  sim.gdina (2,3), itemfit.sx2 (1,2), din(6)

--------------------------------------------------------------
CDM 2.4 (2013-10-24)   --  CRAN CDM 2.4-9
--------------------------------------------------------------

x included two datasets 'data.dcm' and 'data.ecpe'
x eliminated a recoding error in data.Students

--------------------------------------------------------------
CDM 2.3 (2013-09-06)   --  CRAN CDM 2.3-15
--------------------------------------------------------------

x worked on slight improvements of computation time in din
  and gdina
x included multiple group dataset 'data.mg'
x cleaned CDM manual (thanks to Michaela Mayr)

--------------------------------------------------------------
CDM 2.2 (2013-08-20)   --  CRAN CDM 2.2-20
--------------------------------------------------------------

x accelerated rowMaxs function
x included Q-matrix validation function (de la Torre, 2008)
x included joint maximum likelihood estimation and
  classification function for the mixed DINA/DINO model
  -> function din.deterministic
x included another example (based on attribute vectors) 
  in sim.din function
x included polytomous GDINA model (Chen & de la Torre, 2013)  

--------------------------------------------------------------
CDM 2.01 (2013-07-27)   --  CRAN CDM 2.01-11
--------------------------------------------------------------

x fixed a bug in gdm in multiple group estimation: group
  labeling was incorrect and is now included in output group.stat
  and displayed in summary.gdm
x improved convergence behavior for multiple group analysis
x accelerated calculation in gdm: skillspace reduction

--------------------------------------------------------------
CDM 1.5 (2013-07-09)   --  CRAN CDM 1.5-12
--------------------------------------------------------------

x allowed for centering of intercepts and slopes in gdm
x included satured skill distribution estimation in gdm
x fixed a bug in the calculation of RMSEA itemfit, although
  is not clear if the correct formula is used
x fixed a bug in gdm with varying number of item categories
  (thanks to Artur Pokropek)
x included the multidimensional latent class IRT model 
  (Bartolucci, 2007) in gdm (skillspace="est")
x speeded calculation of modelfit.cor

--------------------------------------------------------------
CDM 1.4 (2013-06-30)   --  CRAN CDM 1.4-16
--------------------------------------------------------------

x eliminated a bug in GDINA function for link="log"
x included the parametrization of the reduced RUM model in gdina
x worked on higher order GDINA model which has the argument
  HOGDINA in gdina. Now there is dependency on the fa factor 
  analysis function in the psych package.
x included log-likelihood extractor functions
x included group-wise calculation of RMSEA statistics
  see itemfit.rmsea and the gdm output (thanks to Artur Pokropek)
x (hopefully) fixed a bug in gdm function with differing number of
  item categories (thanks to Artur Pokropek)  
x use frequency patterns in gdm to speed up calculations
  for large data sets (argument 'use.freqpatt' in gdm function)
x included more than one step in gdm function
  - arguments msteps (set to 8 M steps as the default) and
    convM

--------------------------------------------------------------
CDM 1.3 (2013-05-15)   --  CRAN CDM 1.3-17
--------------------------------------------------------------

x included the general diagnostic model (function gdm)
x included a function for calculating equivalent skill classes
  for the DINA/DINO model
x modelfit and anova functions are also accessible for the GDM
x fixed a bug in the function for classification accuracy
x included the RMSEA item fit index
x fixed a bug in gdina estimation
x included coef extractor function

--------------------------------------------------------------
CDM 1.2-3 (2013-02-12)
--------------------------------------------------------------

x included a model fit function modelfit.cor(.din) for the
  comparison of observed and expected correlations
x fixed a bug in the gdina function when applying the model
  with one skill

--------------------------------------------------------------
CDM 1.2-1/1.2-2 (2013-02-01)
--------------------------------------------------------------

x allow for user specified reduced skill space designs
x included tetrachoric correlations in summary.gdina function
x handled zero probability classes in gdina function
x allowed for output and input of attribute patterns in sim.din
  function
x included relative deviance change as convergence criterion in din
  function
x included function for determining a statistically equivalent 
  DINA model (von Davier, 2013)
x changed structure of din summary
x included an anova function
x included a function for estimating classification reliability
x fixed a bug in gdina function when creating design matrices

--------------------------------------------------------------
CDM 1.2.0 (2012-09-30)
--------------------------------------------------------------

x implemented the latent skill space reduction according
  to Xu and von Davier (2008)
x fixed a bug in calculation of estimated parameters
x fixed a bug in IDI calculation
x corrected MLE based classification in case of latent
  classes with a priori spceified zero probability in the
  din function (yet to be done in gdina)
x fixed a bug in standard error calculation in gdina
x speed improvements in din and gdina function
  - included a C function for calculating the posterior 
    distribution
x check of admissible rule in gdina
x include rule "GDINA2" and "GDINA1" ("GDINA1" is 
  equivalent to "ACDM")
x Simulation function for the GDINA model included
x multiple group gdina function for estimation of
  generalized DINA models (de la Torre, 2011)
x new function summary.gdina

--------------------------------------------------------------
CDM 1.1 (2012-07-29)
--------------------------------------------------------------

x highest argument deleted from print.din method
x new argument for summary.din n.skill.classes
x skill pattern probabilities deleted from print.din
x skill pattern summary replaced by n most frequent skill 
  classes in summary.din
x (argument ask=TRUE in summary.din placed by overwrite=FALSE)
x use a different item discrimination index (Lee et al., 2012)
x remove GDINA estimation from this package version because
  it is in an experimental status 
    -> gdina will be included in CDM >= 1.1-1
x new optional attribute for q-matrix specification list of skill 
  labels
x changed ouput in din function
   -> pattern, like and posterior are sorted according to the original 
      input data frame
x remove a bug for models with only attribute
x new argument for din method zeroprob.skillclasses
x new argument for din method equal.itempars
x new arguments for summary.din method log.file and ask
x minor bug fixes in print.din method
x replace internal method rowProds with faster rowProds2
x option for writing an output file of model results included in 
  function summary.din 
