- Change - update
simGDINAto handle CDMs with attribute structures - Added - add an argument
SEfor MCmodel function - Added - add nloptr for optimization
- Added - SISM model is available using the
GDINAfunction andsimGDINAfunction - Change - BUG-DINO model was revised in the
GDINAfunction - Change - examples
- Change - update the package to remove the dependence on nloptr package
- Change - data set: sim10MCDINA
- Change -
GDINAfunction can accept factor or char vector forgroupargument - Change - forward search algorithm implemented in
diffunction now - Change -
Qvalfor Q-matrix validation using an iterative and dynamic procedure - Added - data set: sim10MCDINA2
- Fixed -
CAfunction for models with attribute structures - Fixed -
GDINAfunction when one set of initial values was used (Thanks to Hulya Yigit) - Fixed - Mord statistic calculation in
modelfitfunction when attribute structure exists (Thanks to Haiymiao Yuan)
- Changed - add reference to the JSS article
- Changed - clean references
- Changed - data set format so the package can be used for R versions prior to 3.5
- Added - mesa plot for validating Q-matrix based on sequential G-DINA model
- Fixed - A bug in
GMSCDMfunction (Thanks to Ouyang Xueyuan) - Changed - examples and documentations of
GDINAfunction - Changed - argument of
simGDINAfunction
- Added - Q-validation and model comparison when joint attribute model is not saturated
- Fixed - absolute fit statistics disappear in GUI (since version 2.7.0)
- Fixed - error in calculating CAIC for GDINA estimation (since version 2.7.0)
- Added - MC-DINA model can be estimated using
MCmodelfunction - Added - more components can be extracted using
extractfunction - Fixed -
Qvalfor sequential model using Wald test (since version 2.5.2) - Added - print CAIC and SABIC indices for GDINA estimations
- Changed - bootstrap SEs can be calculated for models with structured attributes
- Added - convergence is checked for
GDINAfunction - Changed - Bonferroni method is replaced by Holm–Bonferroni method as default settings for multiple comparisons in
dif,itemfitandmodelcomp
- Fixed - a mistake in BIC calculation (since version 2.5.2)
- Fixed - M2 calculation in
modelfitwas incorrect for LLM and RRUM - Changed - M2 in
modelfitcan be calculated for higher-order or independent models of joint attribute distribution
- Added -
ICLA()function for attribute profile estimation - Changed - data are aggregated before model estimation for single group CDMs
- Deleted - exported internal functions are removed
- Fixed -
print.GDINA()` prints valid number of individuals by default now - Fixed - covariance matrix cannot be calculated when some individuals are deleted
- Fixed -
Qvalnot work when estimated number of individuals in some latent classes are 0 - Fixed -
extractonly gives valid data
- Added - Predicted cutoff based on the data for Q-matrix validation
- Changed - add .5 for elements 0 when calculating log odds statistic for item fit
- Fixed - model fit cannot be printed in GUI when M2 cannot be calculated
- Added - Attribute Hierarchical structure for A-CDM, LLM and RRUM
- Added - Estimating generalized multiple strategy CDMs using
GMSCDM - Changed -
att.strargument of theGDINAfunction has been updated - Changed - several examples in
GDINAfunction - Fixed - The number of parameters is not correct when attribute strcture exists
- Changed - set anchor attributes for multiple group higher-order CDMs
- Changed - change arguments for
diffunction - Changed - print estimated mixing proportions in R GUI
- Changed - GUI has been largely improved
- Changed - update
modelcompfunction to provide selected models directly - Fixed - calculation of the number of parameters when parameters of some items are fixed
- Fixed - mesa plot with best q-vectors when some q-vectors have the same PVAF
- Fixed - bug when calculating standard error using complete information based on multiple group models
- Changed - arguments for
plotfunction - Added - additional example for
ecpedata - Changed - graphical interface
- Added - experimental functions for simulating and estimating diagnostic tree model (to be optimized)
- Fixed - bugs in
diffunction andstartGDINAfunction - Changed -
logLikfunction to be consistent with default S3 methods - Added - Q-matrix validation using stepwise Wald test
- bug fixed
- This is a major update including a large number of new features. The
GDINAfunction has been largely rewritten for both flexibility and speed. Users are now allowed to fitMS-DINAmodel,Bugsmodels, and define models by providing design matrix and link functions. For joint attribute distribution, in addition to saturated and higher-order models, users are now allowed to fit independent model and loglinear model. Code for joint attribute distribution modelling has been restructured as well. Other major updates include model fit evaluation using M2 statistics and other limited information measures, item-level model selection using likelihood ratio test and score test, classification accuracy evaluation indices, bootstrap standard error estimation, etc. - Note that due to the major updates, some results from this version may be slightly different from those using the previous versions.
- Fixed - a bug in
modelcompin the version 1.4.1 - Changed - adjusted p values are provided for DIF detection
- Changed - update GUI interface
startGDINAto report absolute fit statistics - Changed - estimation methods for higher-order models were adjusted
- Changed - priors can be imposed to structural paprameters in
higher-ordermodels - Changed - missing values were identified when calculating likelihood
- Changed -
GDINAfunction arguments were modified - Changed - outputs for
npar,AIC,BIC,logLikanddeviancefunctions are modified - Changed -
anovafunction is changed and multiple models can be comparied - Changed -
printfunction for the GDINA function is changed - Changed -
summaryfunction is changed - Changed - objects returned from
extractfunction are not rounded - Fixed - bug when extracting covariance matrix and discrimination index (thanks to Kevin Carl Santos)
- Fixed - bug in
diffunction when DIF items are specified - Fixed - bug for calculating standard errors for sequential models
- Fixed - bug - recover the random seeds after running
GDINAanditemfit - Added - standard errors using complete information matrix
- Added - output for
initial.catprobinextractfunction - Added - function
LC2LGto find equivalent latent groups - Added - function
scoreto find score functions - Added - Standard error estimation based on complete information is available
- Fixed - bugs in model estimation with user specified structures
- Fixed - bugs in
summary.GDINAfunction for multiple group estimation - Changed - include the pseudo q-vector 0 for the mesa plot
- Changed - prior distribution is not re-calculated after likelihood calculation in
GDINAfunction - Changed - output for
att.priorinextractfunction - Changed - print number of group for
GDINAfunction - Changed - documents in
simGDINAandGDINA - Changed - examples in
GDINA - Added - extract
ngroupusingextract.GDINAfunction
- Fixed - infinite value issue during M-step optimization
- Fixed -
itemfitfunction for missing data - Fixed - adding monotonic constraints for the sequential models
- Fixed - the maximum likelihood estimation of person attribute through
personparm - Changed - package imports, and suggests
- Changed - slsqp as the first optimizer used for monotonic G-DINA
- Changed - Non zero prior probabilities
- Initial release