Changelog

Change - update simGDINA to handle CDMs with attribute structures
Added - add an argument SE for MCmodel function
Added - add nloptr for optimization
Added - SISM model is available using the GDINA function and simGDINA function
Change - BUG-DINO model was revised in the GDINA function
Change - examples

Change - data set: sim10MCDINA
Change - GDINA function can accept factor or char vector for group argument
Change - forward search algorithm implemented in dif function now
Change - Qval for Q-matrix validation using an iterative and dynamic procedure
Added - data set: sim10MCDINA2
Fixed - CA function for models with attribute structures
Fixed - GDINA function when one set of initial values was used (Thanks to Hulya Yigit)
Fixed - Mord statistic calculation in modelfit function when attribute structure exists (Thanks to Haiymiao Yuan)

Changed - data set format so the package can be used for R versions prior to 3.5

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 MCmodel function
Added - more components can be extracted using extract function
Fixed - Qval for 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 GDINA function
Changed - Bonferroni method is replaced by Holm–Bonferroni method as default settings for multiple comparisons in dif, itemfit and modelcomp

Fixed - a mistake in BIC calculation (since version 2.5.2)
Fixed - M2 calculation in modelfit was incorrect for LLM and RRUM
Changed - M2 in modelfit can 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 - Qval not work when estimated number of individuals in some latent classes are 0
Fixed - extract only 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

Changed - set anchor attributes for multiple group higher-order CDMs
Changed - change arguments for dif function
Changed - print estimated mixing proportions in R GUI
Changed - GUI has been largely improved
Changed - update modelcomp function 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

Added - experimental functions for simulating and estimating diagnostic tree model (to be optimized)
Fixed - bugs in dif function and startGDINA function
Changed - logLik function to be consistent with default S3 methods
Added - Q-matrix validation using stepwise Wald test

This is a major update including a large number of new features. The GDINA function has been largely rewritten for both flexibility and speed. Users are now allowed to fit MS-DINA model, Bugs models, 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.

Changed - update GUI interface startGDINA to report absolute fit statistics
Changed - estimation methods for higher-order models were adjusted
Changed - priors can be imposed to structural paprameters in higher-order models
Changed - missing values were identified when calculating likelihood
Changed - GDINA function arguments were modified
Changed - outputs for npar, AIC,BIC,logLik and deviance functions are modified
Changed - anova function is changed and multiple models can be comparied
Changed - print function for the GDINA function is changed
Changed - summary function is changed
Changed - objects returned from extract function are not rounded
Fixed - bug when extracting covariance matrix and discrimination index (thanks to Kevin Carl Santos)
Fixed - bug in dif function when DIF items are specified
Fixed - bug for calculating standard errors for sequential models
Fixed - bug - recover the random seeds after running GDINA and itemfit
Added - standard errors using complete information matrix
Added - output for initial.catprob in extract function
Added - function LC2LG to find equivalent latent groups
Added - function score to 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.GDINA function 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 GDINA function
Changed - output for att.prior in extract function
Changed - print number of group for GDINA function
Changed - documents in simGDINA and GDINA
Changed - examples in GDINA
Added - extract ngroup using extract.GDINA function

Fixed - infinite value issue during M-step optimization
Fixed - itemfit function 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

GDINA 2.9.3