Supplementary functions for item response models aiming
to complement existing R packages. The functionality includes among others
multidimensional compensatory and noncompensatory IRT models
If you use
sirt and have suggestions for improvement or have found bugs, please email me at [email protected].
The official version of
sirt is hosted on CRAN and may be found here.
The CRAN version can be installed from within R using:
The version hosted here is the development version of
The GitHub version can be installed using
----------------------- sirt NEWS -----------------------------
sirt: Supplementary Item Response Theory
Alexander Robitzsch IPN - Leibniz Institute for Science and Mathematics Education at Kiel University
Questions or suggestions about sirt can be sent to: [email protected]
For reporting a bug, please provide always a reproducible R script and (if necessary) a corresponding dataset.
ADDED * included optimization function for invariance alignment which minimizes number of DIF effects in invariance.alignment() by setting 'align.pow=c(0,0)'. This optimization function corresponds to the L0 penalty. NOTE * included a warning message in polychoric2() if lowest category is larger than zero (due to correspondence with Njaal Foldnes) NOTE * included argument 'use_pbv' in polychoric2() which allows use of pbv package for computation of bivariate normal probabilities NOTE * included arguments 'conv', 'rho_init' and 'weights' in polychoric2(). Exported Rcpp-level function sirt_rcpp_polychoric2(). FIXED * fixed a bug when computing standard errors in lsem.estimate() when arguments 'pseudo_weights>0' or 'use_lavaan_survey=FALSE' were used (thanks to Gabriel Olaru) NOTE * changed default in lsem.estimate() of argument 'sufficient_statistics' to TRUE. This option can only be used for continuous data. ADDED * included option for sampling weights in lsem.estimate() and lsem_local_weights() ADDED * included uniform and Epanechnikov kernel for LSEM estimation by including the argument 'kernel' in lsem.estimate() ADDED * LSEM can now be additionally estimated in a joint estimation approach ('est_joint=TRUE'). Invariance constraints on parameters are allowed by 'par_invariant' in lsem.estimate() NOTE * included check on invariance constraint for fixed parameters in lsem.estimate() (thanks to Ingo Seifert) NOTE * excluded computation of standard errors in lsem.estimate() if 'est_joint=TRUE' is used (thanks to an example of Ingo Seifert) NOTE * now remove cases with missing values on moderators in lsem.estimate() (thanks to @Deleetdk, issues #8, #9) FIXED * fixed a bug in xxirt() caused by incorrect data processing which stopped the function (thanks to Hyejin Shim) NOTE * fit initial model in lsem.estimate() if sufficient statistics are used based on pairwise deletion (thanks to Martina Bader) ADDED * added bootstrap functionality for LSEM in the function lsem.bootstrap()
DATA * included/modified datasets: --- EXAMP * included/modified examples: invariance.alignment (1), polychoric2 (1), lsem.estimate (1.1, 1.5)
ADDED * included sampling weights for invariance alignment in function invariance_alignment_cfa_config() FIXED * fixed numerical instabilities in wle.rasch() (thanks to Tom Tinkler) FIXED * fixed a bug in lsem.estimate() when sufficient_statistics=TRUE is applied in case of missing data (thanks to Johanna Hartung) NOTE * lsem.estimate() does not use anymore the lavaan.survey package as the default. The behavior of previous versions can be obtained with the argument value 'use_lavaan_survey=TRUE'. ADDED * added argument 'pseudo_weights' in lsem.estimate() which provides an ad-hoc solution for fitting LSEMs for all lavaan models. FIML and WLSMV is now available in lsem.estimate(), see Examples 2 and 3 (requested by Ulrich Schroeders and Andrea Hildebrandt) ADDED * added utility function lsem_local_weights() for computing local weights for LSEMs ADDED * added argument 'judge' for judge identifiers in btm() which allows computation of fit statistics for judges in pairwise comparison data (suggested by Jem Stach) FIXED * fixed a bug in estimation of home advantage parameter in btm() ADDED * added simulation function btm_sim() for simulating data from the extended Bradley-Terry model NOTE * removed argument 'seed' from class.accuracy.rasch()
DATA * included/modified datasets: --- EXAMP * included/modified examples: invariance.alignment (7), lsem.estimate (1,2,3), btm (3,4)
NOTE * changes in estimation algorithm in noharm.sirt(). Disabled estimation based on tetrachoric correlations and, hence, removed argument 'modesttype'. Included arguments 'optimizer' and 'par_lower'. ADDED * included arguments 'pm' and 'N' for product-moment matrix input in noharm.sirt() which allows input of a product-moment matrix ADDED * added additional estimation methods 'MED', 'LTS', 'HUB', 'L1' and 'L0' in linking.haberman() NOTE * included utility function linking_haberman_itempars_prepare() for generating 'itempars' argument for linking.haberman() NOTE * included utility function 'L0_polish' NOTE * included automatic functionality for detecting linking of 1PL models in linking.haebara() NOTE * included example for 2PL copula model (?rasch.copula, Example 1; requested by Jarkko Hautala) ADDED * included utility functions invariance_alignment_cfa_config() and invariance_alignment_simulate() for invariance alignment (due to a discussion with Emily Harris)
DATA * included/modified datasets: --- EXAMP * included/modified examples: noharm.sirt (1,2.3,3), linking.haberman (6,7,8), rasch.copula2 (1), invariance.alignment (7)
NOTE * complete restructure of invariance.alignment() function which now uses R internal optimizers stats::optim() or stats::nlminb(). Included postprocessing function invariance_alignment_constraints(). NOTE * added utility functions sirt_matrix2() and several sirt_colFoo()'s NOTE * renamed eigenvalues.sirt() into sirt_eigenvalues() NOTE * included argument 'adjust_main_effects' in linking.haberman() ADDED * included linking method linking.haebara() for linking multiple groups with robust and non-robust estimation function NOTE * included more arguments in summary.rasch.pairwise(), summary.rasch.evm.pcm() NOTE * changed default of 'theta' argument in lsdm() function ADDED * included alternative optimization method of absolute value loss function in lsdm() and added arguments 'wgt_theta' and 'distance' NOTE * removed dependency in lsdm() function from ic.infer::orlm() NOTE * included utility functions sirt_sum_norm(), sirt_dnorm_discrete(), sirt_abs_smooth(), sirt_permutations() NOTE * included a plot method for lsdm() NOTE * removed dependency in rasch.copula2() from gtools::permutations()
DATA * included/modified datasets: --- EXAMP * included/modified examples: invariance.alignment (6), linking.haberman (1,5), linking.haebara (1), lsdm (1,5,6), sirt-utilities (5,6)
NOTE * included estimation details of btm() function
(requested by San Verhavert)
NOTE * included argument 'center' in rasch.jml() which now also allows
option for centering item difficulties
(request thoughtafter, issue #3)
FIXED * exclude non-converged model estimates for permutation datasets
in lsem.permutationTest() (reported by Moritz Breit)
NOTE * added argument 'save.gamma.testlet' in mcmc.3pno.testlet()
(requested by Vladimir Volokhonsky, issue #4)
NOTE * corrected internet address for DIMTEST software in ?conf.detect (thanks to Julius Fenn) FIXED * fixed numerical problems in person.parameter.rasch.copula() (detected by CRAN checks of examples for r-devel-windows-ix86+x86_64; https://cran.r-project.org/web/checks/check_results_sirt.html) NOTE * changed column 'itemdiff.correction' in output 'item' in rasch.jml() in case of constrained item parameters (noted by thoughtafter, issue #5)
DATA * included/modified datasets: --- EXAMP * included/modified examples: rasch.jml (1)
FIXED * fixed a bug in tam2mirt() for models with polytomous data which do not contain any dichotomous item (thanks to Sarah Anne Aretz) FIXED * fixed an incorrect subject order in outfit/infit statistics in btm() (thanks to Stephen Holmes) FIXED * fixed a bug in btm() with 'eps=0' for persons with extreme scores (thanks to SanVerhavert) FIXED * completely rewritten estimation algorithm of rm.sdt() which produced biased estimates in previous sirt versions (pointed out by Jonathan Lehrfeld, Kyosuke Bunij, Satoshi Usami). NOTE * added graded response model at item level for signal detection model for rating data in rm.sdt() (suggested by Jonathan Lehrfeld) NOTE * included utility function sirt_optimizer()
DATA * included/modified datasets: --- EXAMP * included/modified examples: ---
NOTE * changed matrix types in rm.facets() Rcpp data processing functions. This addresses UBSAN problems causing detected problems in rm.facets(): [...] Rcpp/include/Rcpp/internal/caster.h:30:25: runtime error: nan is outside the range of representable values of type 'int'. SUMMARY: UndefinedBehaviorSanitizer: undefined-behavior in [...]/data/gannet/ripley/R/test-clang/Rcpp/include/Rcpp/internal/caster.h:30:25 FIXED * fixed a bug in tam2mirt() due to changed output structure in mirt package (thanks to Sarah Anne Aretz)
DATA * included/modified datasets: --- EXAMP * included/modified examples: ---
FIXED * fixed a bug in postprocessing in rm.sdt() function FIXED * fixed bugs in mcmc_XXX() functions when mcmc object (from coda) is a list FIXED * handle case of characters "[", "]", "," in variable names in functions mcmc_derivedPars() and mcmc_WaldTest() FIXED * mcmc_WaldTest() now also handle cases of multicollinear hypotheses FIXED * handled NaN problems in noharm.sirt() NOTE * included argument 'reliability' in noharm.sirt() NOTE * changed argument 'logfile' to 'file' in summary.noharm.sirt() NOTE * rewrote some Rcpp function because of memory problems documented at https://cran.r-project.org/web/checks/check_issue_kinds.html
DATA * included/modified datasets: --- EXAMP * included/modified examples: ---
FIXED * fixed a bug in rm.facets() when argument rater_item_int=TRUE was used NOTE * exported function rm_proc_data() for processing rating datasets FIXED * avoided numeric overflow in pcm.fit() (thanks to Chen Chen) FIXED * fixed broken function lsdm() because exported functions in dependent package 'ic.infer' have been changed
DATA * included/modified datasets: --- EXAMP * included/modified examples: ---
NOTE * included different parameter centering options in rm.facets() FIXED * fixed a bug in data processing in xxirt() function NOTE * included EM acceleration method 'PEM' in rm.sdt() and rm.facets() NOTE * corrected example for Q3() function (thanks to Amina Fraij) ADDED * added prior distributions for item parameters in rm.sdt() NOTE * included option for defining a reference rater in rm.facets(). See Example 2.6. NOTE * included argument 'rater_item_int' for modelling rater-item-interaction effects in rm.facets() (see Example 2.7) NOTE * included more arguments for starting values in rm.sdt() NOTE * exported utility function sirt_attach_list_elements() and several sirt_summary_print_XXX() functions ADDED * included synthetic dataset 'data.befki' as an example for local structural equation modeling (thanks to Ulrich Schroeders, Stefan Schipolowski & Oliver Wilhelm)
DATA * included/modified datasets: data.befki, data.befki_resp EXAMP * included/modified examples: Q3 (1), rm.facets (2.6, 2.7)
ADDED * added option for additive linking of item slopes in linking.haberman() FIXED * fixed broken mcmc.3pno.testlet() and mcmc.2pno() functions (thanks to Hyeon-Ah Kang) NOTE * slight changes in output in MCMC IRT functions like mcmc.3pno.testlet(). Included argument 'file' in summary.
DATA * included/modified datasets: --- EXAMP * included/modified examples: linking.haberman (1)
NOTE * included synthetic datasets simulated from original datasets from the test theory book of Eid and Schmidt (2014) NOTE * moved packages gtools, ic.infer, igraph, Matrix, pbivnorm, psych and sm from 'Imports' to 'Suggests'. Removed combinat, MASS and plyr packages from 'Imports'. NOTE * exported sirt_rbind_fill() utility function which is a reimplementation of plyr::rbind.fill() NOTE * exported utility functions sirt_fisherz() and 'sirt_antifisherz() NOTE * fixed potential problems with frequencies of zero in dif.logistic.regression() (thanks to Armin Haba)
DATA * included/modified datasets: data.eid.kap4, data.eid.kap5, data.eid.kap6, data.eid.kap7 EXAMP * included/modified examples: ---
NOTE * moved amh(), loglike_mvnorm(), mlnormal() and pmle() functions to LAM package (soon at CRAN) NOTE * moved starts_cov() and 'starts_sim1dim' functions to STARTS package (soon at CRAN) FIXED * fixed a bug in IRT.likelihood() for fitted models in mirt which occured due to an output restructure in mirt FIXED * fixed a bug in mirt.specify.partable() NOTE * avoid division by zero in IRT.mle() which caused numerical problems (thanks to Marit List) NOTE * some subfunctions in xxirt() were rewritten in Rcpp for reducing computation time NOTE * removed dependency from MCMCpack package ADDED * added functionality for handling missing values unter MCAR assumption in lsem.estimate(). The argument 'sufficient_statistics=TRUE' allows for local estimation of SEM based on weighted means and weighted covariances. (suggested by Ulrich Schroeders) FIXED * fixed a bug in lsem.estimate() if the object name of the argument 'lavmodel' is not 'lavmodel' (thanks to Ulrich Schroeders and Jenny Wagner)
DATA * included/modified datasets: --- EXAMP * included/modified examples: lsem.estimate (1)
NOTE * included option for choosing lavaan::lavaan() or lavaan::sem() as estimation function in function lsem.estimate(). In previous versions, only the lavaan::sem() function could be chosen (thanks to Jenny Wagner). NOTE * restructured Rcpp functions NOTE * removed CallSwitch() function from namespace and package NOTE * removed minor bug in labeling items in sim.raschtype() FIXED * fixed a bug in smirt() with handling of parameter bounds which was recently introduced in sirt 1.13 (thanks to David Armstrong)
DATA * included/modified datasets: --- EXAMP * included/modified examples: ---
NOTE * changed output of 'sim.raschtype' which is now of class data frame NOTE * added a not in 'lsem.estimate' that the model syntax should be always specified in an object lavmodel (thanks to a communicated problem of Jenny Wagner) NOTE * included utility function 'mcmc_Rhat' for computation of the Rhat statistic from a single chain NOTE * added statistics of average conditional covariances and MADCOV in 'conf.detect' and 'expl.detect'. An example for DETECT analysis for simulated unidimensional data has been included (Example 3) which shows that DETECT can become negative (motivated by discussions with Berend Terluin, Weiyi Cheng, Yifan Liu and Yunfei Chang) NOTE * added summary method for 'conf.detect' NOTE * disabled default fixed seed in 'expl.detect'
DATA * included/modified datasets: --- EXAMP * included/modified examples: mcmc_Rhat (1), conf.detect (3)
FIXED * fixed a bug in 'conf.detect' and 'expl.detect' with too few items (routine crashed, thanks to Berend Terluin) FIXED * fixed a bug in data preprocessing in 'mlnormal' function NOTE * included reference to subscore package in 'prmse.subscores.scales' NOTE * extended output values in 'categorize' function NOTE * added arguments for boundary values of a, b, c and d parameters in noncompensatory IRT model function 'smirt' (requested by Chia-Ling Hsu) ADDED * added robust linking method for multiple studies based on bisquare robust regression estimation in 'linking.haberman'. In addition, a summary method has been added. NOTE * included functions 'starts_cov' and 'starts_sim1dim' for the STARTS model of Kenny and Zautra
DATA * included/modified datasets: --- EXAMP * included/modified examples: mlnormal (3), linking.haberman (5), starts_cov (1)
NOTE * included utility function 'prior_model_parse' for convenient specifications of priors for the 'amh' function NOTE * included utility function 'loglike_mvnorm' for calculation of log-likelihood for a multivariate normal distribution based on sufficient statistics ADDED * added deterministic optimization for Bayesian model estimation in the function 'pmle' (penalized maximum likelihood estimation). See the 'amh' help page. NOTE * fixed a bug in LSEM plot function for subsets of parameters (thanks to Johanna Hartung) NOTE * corrected a bug in Example 1 in 'equating.rasch' (thanks to Kelvin Gregory) NOTE * changed labelling of response pattern in 'md.pattern.sirt' to integers NOTE * included example in 'nedelsky-methods' for fitting the Nedelsky model with xxirt function NOTE * added history of adaptive proposal standard deviations in 'amh' function in output NOTE * added sampled parameter vector with highest value of posterior distribution in 'amh' for comparison with 'pmle' function NOTE * added comparison of MAP, multivariate MAP and mean estimator in 'amh' in value 'comp_estimators'. Now also included in 'summary.amh'. NOTE * Defunct functions: 'rasch.conquest', 'rasch.pml2' , 'yen.q3', 'testlet.yen.q3' NOTE * included utility functions 'CallSwitch' and 'parmsummary_extend' NOTE * included utility functions 'pow', 'tracemat', 'soft_thresholding', 'hard_thresholding', 'bounds_parameters', 'dimproper', 'ginverse_sym' (see '?sirt-utilities') ADDED * included estimation routine 'mlnormal' for (restricted) maximum likelihood estimation for mean and covariance structures for a multivariate normal distribution (e.g. multilevel models, factor analysis, structural equation models, multilevel structural equation models, social relations model, ...). Prior distribution and lasso penalities are also accomodated.
DATA * included/modified datasets: --- EXAMP * included/modified examples: prior_model_parse (1), loglike_mvnorm (1), amh (1), equating.rasch (1), amh (1), nedelsky-methods (1), sirt-utilities (1,2,3,4), parmsummary_extend (1), mlnormal (1,2)
FIXED * fixed a bug in 'mirt.wrapper.coef' which occured due to changes in output structures of mirt package ADDED * added user defined item response model estimation 'xxirt' with corresponding S3 methods ADDED * added experimental adaptive Metropolis-Hastings routine 'amh' with corresponding S3 methods ADDED * added utility functions for objects of class 'mcmc.list' (from coda package): 'mcmc_coef', 'mcmc_vcov', ..., 'mcmc_WaldTest' NOTE * included inverse gamma distribution ('rinvgamma2') with parametrization in prior sample size and prior variance
DATA * included/modified datasets: --- EXAMP * included/modified examples: rasch.jml (1), xxirt (1,2), xxirt_createParTable (1), xxirt_createThetaDistribution (1), amh (1,2), mcmc_coef (1), rinvgamma2 (1)
FIXED * fixed a bug in summary function of MCMC based item
response models (thanks to Andres Burga)
FIXED * fixed a bug in 'rasch.mml2' when using the 'center.b' argument (thanks to Sarah Thomas) NOTE * renamed functions 'yen.q3' into 'Q3' and 'testlet.yen.q3' into 'Q3.testlet'. The old function names can still be used. NOTE * included argument 'standardized' in 'lsem.estimate' which adds standardized solutions in the output using 'lavaan::standardizedSolution' (thanks to Janina Kuenecke) NOTE * cleaned namespace file and defined imported functions in a more explicit way
DATA * included/modified datasets: --- EXAMP * included/modified examples: lsem.estimate (1)
NOTE * allows thinning in 'plot.mcmc.sirt' function NOTE * fixed a minor bug in summary output of 'rasch.mml2' for irtmodel='missing1' NOTE * included argument 'zerosum' in 'rasch.pairwise' and 'rasch.pairwise.itemcluster' to apply a zero sum constraint of for item difficulties (suggested by Chen-Wei Liu) NOTE * included example for bootstrapping standard errors for 'rasch.pairwise' (suggested by Chen-Wei Liu)
DATA * included/modified datasets: --- EXAMP * included/modified examples: rasch.pairwise (1), dif.logistic.regression (2)
NOTE * corrected a non-working example in 'lsem.estimate' (thanks to Takuya Yanagida) NOTE * changed defaults and outputs in 'mcmc.list.descriptives' and 'plot.mcmc.sirt' ADDED * included argument 'est.delta' in 'rasch.mml2' for item specific estimation of delta parameters in irtmodel='missing1'. Now, the specification of est.delta has slightly been changed. ADDED * included function 'likelihood.adjustment' for adjusting individual likelihood NOTE * disabled the broken plot method for 'noharm.sirt' objects
DATA * included/modified datasets: --- EXAMP * included/modified examples: lsem.estimate (1), rasch.mml2 (7.12-14, 8), likelihood.adjustment (1)
ADDED * included function 'btm' for estimating an extended Bradley-Terry model (suggested and fixed by Chris Weadon) NOTE * added minimum and maximum values in summary of 'lsem.estimate' NOTE * included JML estimation of Bradley-Terry model ('btm') with elimination of persons with extreme scores NOTE * included infit and outfit statistics in 'btm' NOTE * changed layout of summary of 'lsem.estimate' and 'plot.lsemPermutationTest' NOTE * added separation index for Bradley-Terry model (suggested by Sven De Maeyer) FIXED * fixed a bug in 'categorize' function
DATA * included/modified datasets: data.pw01 EXAMP * included/modified examples: btm (1,2)
NOTE * added arguments 'minnode' and 'maxnode' in 'R2conquest' which now allows full specification for the theta grid used in numerical approximation FIXED * fixed a bug in 'truescore.irt' when no standard errors were provided ADDED * added utility functions 'categorize' and 'decategorize' for categorizing variables in a data frame ADDED * added function 'lsem.estimate' for fitting local structural equation models (LSEM). In addition, the permutation test is implemented in 'lsem.permutationTest'.
DATA * included/modified datasets: data.lsem01 EXAMP * included/modified examples: categorize (1,2), lsem.estimate (1)
NOTE * changed default for argument 'itemcodes' in R2conquest fuction NOTE * included S3 methods for probabilistic Guttman model 'prob.guttman' ADDED * included plot method for hierarchical rater model 'rm.sdt' NOTE * changed calculation of transformed parameters in 'rm.sdt' FIXED * fixed a bug in 'rm.sdt' for datasets with differing number of item categories ADDED * added option for specifying a discrete latemt variable in 'rm.sdt' function. Options for fixed item parameters were also added. NOTE * added arguments for bounds of item slope and item difficulty parameters in 'rasch.mml2' (requested by Aysenur Erdemir) ADDED * added option for estimating the missingness delta parameter in 'rasch.mml2' and 'irtmodel="missing1"' FIXED * fixed a bug in 'rasch.mml2' for missing data models and included CALL in summary FIXED * corrected missing output of individual likelihood and posterior in 'rasch.mml2' for irtmodel='missing1'
DATA * included/modified datasets: data.bs07a EXAMP * included/modified examples: data.eid (4.3a), data.read (8e,9a), data.bs (07a), rm.sdt (2,3), rasch.mml2 (1,7)
NOTE * defined several S3 methods for 'rasch.copula2' and 'rasch.copula3' NOTE * adapted S3 methods for mirt because the new class 'SingleGroupClass' is defined in mirt >= 1.7 and removed 'ConfirmatoryClass' and 'ExploratoryClass' methods NOTE * exported the object pcm.fit (it was spuriously non-exported in sirt 1.2; reported by Sven Liepertz) NOTE * included the argument 'ignore.pid' for ignoring person identifiers in 'R2conquest' function (suggested by Kelvin Gregory) NOTE * changed Example 1 in 'f1d.irt' due to a change in Suggests regarding the MASS package in the psych package which caused an error in CRAN checks (suggested by William Revelle) ADDED * included S3 methods logLik, IRT.likelihood, IRT.posterior, IRT.modelfit, summary.IRT.modelfit and IRT.irfprob for functions 'rasch.mml2', 'smirt', 'rasch.mirtlc', 'gom', 'rm.facets' and 'rm.sdt' NOTE * extended modelfit.sirt for objects of class gom ADDED * included function 'modelfit.cor.poly' for assessing absolute model fit for polytomous item responses ADDED * included arguments for fixing item and rater parameters in 'rm.facets' ADDED * included IRT.factor.scores method for 'rm.facets' and 'rm.sdt' ADDED * included function 'IRT.mle' for estimating MLE, WLE and MAP factor scores with user defined item response functions
DATA * included/modified datasets: --- EXAMP * included/modified examples: f1d.irt (1), rasch.mml2 (1), rasch.mirtlc (1), gom.em (1), rm.facets (2), IRT.mle (1,2,3)
NOTE * included standard errors of WLE parameters in 'wle.rasch'
NOTE * corrected formulas help file for 'pcm.conversion'
(thanks to Thomas Kiefer)
FIXED * fixed computational problems for calculating tetrachoric
correlation with empty cells in 'tetrachoric2' resp. 'polychoric2' (method of Olsson; thanks to a dataset privided by Ulrike Kipman) NOTE * included S3 methods IRT.likelihood, IRT.posterior and IRT.irfprob for two classes (ConfirmatoryClass and ExploratoryClass) of mirt objects ADDED * included item and person fit statistic in 'pcm.fit' for the partial credit model (suggested by Hiroyuki Yamada and Sven Liepertz) NOTE * included C-test dataset 'data.g308' from Schroeders et al. (2014) NOTE * included several examples from the book of Eid and Schmidt: "Testtheorie und Testkonstruktion", see data.eid
DATA * included/modified datasets: data.g308, data.eid EXAMP * included/modified examples: pcm.conversion (1), pcm.fit (1,2,3), brm.sim (1), data.g308 (1), data.eid (1,2)
NOTE * included arguments for mean and covariance matrix for
confirmatory multidimensional models in 'f1d.irt'
NOTE * removed 'lavaanify.sirt' from sirt package and
included this function as 'lavaanify.IRT' in
FIXED * fixed a bug in 'lavaan2mirt' (thanks to Hong Giang Pham)
NOTE * extended some person fit statistics to handle datasets with missing responses (thanks to Hiroyuki Yamada) ADDED * included functions for estimating the Beta item response model for continuous item responses
DATA * included/modified datasets: --- EXAMP * included/modified examples: f1d.irt (2), brm.sim (1)
NOTE * included argument 'progress' in 'gom.em' for optional
printing of iteration progress
NOTE * disabled warnings produced in 'rasch.mirtlc' by the glm
FIXED * fixed a bug in calculating the posterior in 'mirt.wrapper'
for nonlinear item response models
FIXED * fixed a bug in running exploratory factor analysis with
one dimension in noharm.sirt (thanks to Russell Zheng)
FIXED * fixed a bug in datawide2long with Q-matrix input
FIXED * fixed a bug in data handling in 'invariance.alignment'
with some missing intercepts and slopes
ADDED * included an extension of lavaan syntax to accomodate
guessing and slipping parameters and "__" (item shortage
declaration) -> function 'lavaanify.sirt'
The function lavaan2mirt also includes these options.
NOTE * included coef and vcov method for function 'rasch.evm.pcm'
NOTE * included argument weights in 'dirichlet.mle'
NOTE * included utility function for conversion of different
parameterizations in partial credit model
NOTE * included helper function 'mirt.wrapper.itemplot'
NOTE * enabled progress argument in rasch.copula functions
NOTE * renamed DeCarlo's hierarchical rater model based on signal detection theory 'rm.hrm' into 'rm.sdt' NOTE * fixed an error in description of missing IRT models in 'rasch.mml2.Rd' (thanks to Carmen Koehler) NOTE * fixed a bug in 'tam2mirt' for polytomous items (thanks to Michal Modzelewski) ADDED * included function 'mirt.specify.partable' for modifying a parameter table in mirt package NOTE * included argument item type for polytomous data in 'lavaan2mirt' NOTE * updated 'lavaanify.sirt' function because of changes in the lavaan::lavaanify function from lavaan 0.5-16 to the lavaan 0.5-17 package NOTE * changed mean constraint for rater parameters in 'rm.facets' ADDED * included estimation and simulation functions for the Nedelsky model ('nedelsky.sim', 'nedelsky.irf')
DATA * included/modified datasets: data.ratings2, data.si06, data.ratings3 EXAMP * included/modified examples: dirichlet.simul (1,2), data.read (2), mirt.model.vars (1), rasch.mirtlc (4), data.long (3,4), lavaanify.sirt (1), lavaan2mirt (1), dirichlet.mle (1,2), pcm.conversion (1), data.big5 (1), mirt.specify.partable (1), nedelsky-methods (1,2)
ADDED * included wrapper function of mirt output of extracting
item parameters ('mirt.wrapper.coef')
EXAMP * added more examples for model estimation in mirt package
using customized item response functions and/or
customized latent structure distributions
o Ramsay's quotient model ('sim.qm.ramsay', Example 1)
o Latent class analysis ('data.read', Model 7)
o Mixed Rasch model ('data.read', Model 8)
o Located latent class model/Latent class IRT model
('data.read', Model 12; 'rasch.mirtlc', Example 4, Model 4)
o Probabilistic Guttman model ('prob.guttman', Example 1,
o Nonparametric item response model with monotone item
response functions ('prob.guttman', Example 4, Model 4)
o Discrete grade of membership model ('gom.em', Simulated Example 2, Model 2m) o Multimensional item response model with discrete traits ('data.read', Model 13)
o DINA model ('data.read', Model 14)
o Rasch copula model ('rasch.copula2', Example 5)
o Rasch model with non-normal distibution ('data.read', Model 15)
ADDED * included wrapper function of mirt output for extracting likelihood, posterior and expected counts ADDED * extended modelfit.sirt for fitted objects from mirt package ADDED * included function 'lavaan2mirt' for converting syntax for the lavaan package to syntax for the mirt package ADDED * included function 'tam2mirt' for converting a fitted object in TAM package into an object of class mirt in the mirt package NOTE * corrected some typos and improved some examples (thanks to Phil Chalmers) EXAMP * included an example for running Mplus from within R using the MplusAutomation package ('data.read', Model 1) NOTE * renamed functionrasch.conquest into R2conquest -> function rasch.conquest can still be used ADDED * included function 'mirt.wrapper.fscores' for computing EAPs, MAPs and MLEs for objects of class mirt
DATA * included/modified datasets: --- EXAMP * included/modified examples: mirt.wrapper (1,2,3,4,5), sim.qm.ramsay (1), modelfit.sirt (1), data.read (1,3,7,8,12,13,14), lavaan2mirt (1,2,3), tam2mirt (1,2,3,4), prob.guttman (1), gom.em (2), equating.rasch (1), data.reck (2), rasch.mirtlc (4), rasch.copula2 (5)
NOTE * used the operator '::' for functions which are used from other packages than sirt in example files for datasets FIXED * fixed a bug in plot.noharm.sirt for unidimensional models NOTE * fixed a minor bug in summary in 'summary.rasch.pml2' NOTE * included argument 'use.maxincrement' for controlling increments in 'rasch.pml3' NOTE * speeded invariance alignment function and choose now different starting values to ensure a global minimum. Changed also default parameters to obtain more stable results. NOTE * excluded plotting function of 'invariance.alignment' NOTE * changed calculation of measures of noninvariance in invariance.alignment. Now, the noninvariance measures in 'invariance.alignment' and 'linking.haberman' are equivalent.
DATA * included/modified datasets: --- EXAMP * included/modified examples: data.read, data.big5, invariance.alignment (4,5)
FIXED * fixed a bug in reliability calculation in noharm.sirt NOTE * included plots for 'noharm.sirt' using the semPlot package ADDED * included first version of statistical implicative analysis ('sia.sirt') ADDED * included function 'rasch.conquest' for running ConQuest from within R
DATA * included/modified datasets: --- EXAMP * included/modified examples: data.mixed1, noharm.sirt (1,2), sia.sirt (1), data.read, rasch.evm.pcm (4)
FIXED * fixed a bug in calculation of marginal parameters for testlet model ADDED * included the partially compensatory item response model (irtmodel='partcomp' in 'smirt') NOTE * speeded 'rasch.copula' functions a bit NOTE * included argument 'increment.factor' in rasch.copula to improve convergence behavior FIXED * fixed a bug in 'rasch.mirtlc' in saved parameter estimates for models with different initial estimates (n.starts>0) ADDED * included function for calculating marginal item parameters for testlet (bifactor) model ('testlet.marginalized') ADDED * extended 'modelfit.sirt' for objects of class tam.mml, tam.mml.2pl and tam.fa (from the TAM package) FIXED * fixed a bug in parsing residual matrix from NOHARM output in 'R2noharm' NOTE * extended output in mcmc.list. descriptives NOTE * changed chi square p value in R2noharm in Values according to summary ADDED * included R implementation of NOHARM ('noharm.sirt') NOTE * modelfit.sirt also allows for objects of class noharm.sirt NOTE * included Divgi estimation method of tetrachoric correlation ADDED * included maximum likelihood estimation of polychoric correlation
DATA * included/modified datasets: --- EXAMP * included/modified examples: data.wide2long (1), smirt (3), modelfit.sirt (1), R2noharm (4), mcmc.list.descriptives (1), noharm.sirt (1,2), data.read, tetrachoric (2), polychoric2 (1)
ADDED * included function 'isop.test' for testing the ISOP model
ADDED * included (non-iterative) eigenvector method for estimating
the partial credit model -> 'rasch.evm.pcm'
NOTE * included argument 'ref.item' in rasch.mirtlc for specification
of reference items
ADDED * included test of de Gooijer and Yuan (2011) that CSN is non-positive
ADDED * included option for testing differential item functioning in rasch.evm.pcm NOTE * changed convergence behavior in rasch.mirtlc for 2PL models ADDED * robust linking alternative to mean-mean linking ('linking.robust') NOTE * corrected output for hierarchical rater model ('rm.hrm') ADDED * included simple latent class model for two exchangeable raters ('lc.2raters') ADDED * extended unidimensional copula model to multiple dimensions ('rasch.copula3') NOTE * changed numerical differentiation parameters in several functions to a smaller value ADDED * included missing data IRT model ('rasch.mml2', irtmodel='missing1') NOTE * included person (containing person parameters) as a value in rasch.copula2 and rasch.copula2 FIXED * fixed a bug in 'testlet.yen.q3' NOTE * allowed correlated residuals in reliability calculation of 'reliability.nonlinearSEM' NOTE * included reliability estimation for pseudo maximum likelihood estimation ('rasch.pml3') NOTE * included marginal item parameters in the output in testlet model ('mcmc.3pno.testlet')
DATA * included/modified datasets: data.reck73C1a, data.reck73C1b, data.reck75C2, data.reck78ExA, data.reck79ExB, data.liking.science, data.si02, data.si03, data.si04, data.si05 EXAMP * included/modified examples: isop.test (1), rasch.evm.pcm (1,2,3,4,5,6), data.reck (2,3), rasch.mirtlc (3,4,5), gom.em (2), unidim.test.csn (1,2), gom.em (1), linking.robust (1,2,3), lc.2raters (1), data.wide2long (2), rasch.copula (4), rasch.mml2 (7), reliability.nonlinearSEM (1)
NOTE * included further values in smirt to allow to apply itemfit.sx2 in the CDM package FIXED * fixed a bug in output of person parameters in smirt (missing standard errors; thanks to Balal Izanloo) NOTE * included empirical item-attribute discriminations in lsdm NOTE * changed NAMESPACE and DESCRIPTION files due to a change in Rcpp FIXED * fixed a bug for constraint in mu for multidimensional models in rasch.mml2 FIXED * fixed warning messages when reading NOHARM output files NOTE * stabilized estimation of tetrachoric correlation coefficients for Tucker method (addition of cell sizes with a value .5) ADDED * included the method of Bonett and Price for calculation a tetrachoric correlation NOTE * included additional outputs and item parameters for the HRM-SDT (thanks to a discussion with Robin Gleeson) NOTE * extended modelfit.sirt to objects of class R2noharm
DATA * included/modified datasets: data.timss07.G8.RUS, data.reck611 EXAMP * included/modified examples: fuzcluster (1), modelfit.sirt (1)
ADDED * included maximum likelihood estimation for discrete data in the belief function framework ('fuzdiscr') ADDED * included pseudo-log-likelihood estimation for fractional item responses (values between 0 and 1) in 'rasch.mml2' ADDED * included clustering fur fuzzy continuous data (belief function framework) ('fuzcluster')
DATA * included datasets: data.pirlsmissing EXAMP * included examples: mcmc.2pno.ml (3), rasch.mml2 (7)
FIXED * fixed a bug in 'data.wide2long' NOTE * changed Rcpp code in eigenvalue calculation due to a CRAN request
DATA * included datasets: data.big5.qgraph EXAMP * included examples: ---
NOTE * included approximating function of expected percent score in truescore.irt NOTE * added a check function for inits in smirt (thanks to Balal Izanloo) FIXED * fixed a bug in output of person parameters (thanks to Balal Izanloo) NOTE * speeded latent regression model 'latent.em.regression.raschtype' ADDED * included latent.em.regression.normal for normally distributed outcomes NOTE * included factor for convergence acceleration in invariance.alignment NOTE * included summary for latent regression models ADDED * included function 'mle.pcm.group' for maximum likelihood estimation in generalized item response models which includes an option for estimating group parameters NOTE * changed algorithm in 'invariance.alignment' into a two-step approach NOTE * included plot method for function 'invariance.alignment'
DATA * included datasets: data.inv4gr, data.ess2005, data.mcdonald.rape, data.si01 EXAMP * included examples: lsdm (4,5), invariance.alignment (3), plausible.value.imputation.raschtype (2,3)
ADDED * included function for automated finding of a scoring key in a data frame with raw item responses ('automatic.recode') FIXED * fixed a bug in linking.haberman function NOTE * added argument personpars in linking.haberman for transformation of person parameters ADDED * included alignment method for approximate invariance (Asparouhov & Muthen, 2013) ADDED * included function truescore.irt for conversion of trait scores into true scores for unidimensional item response models
DATA * included item parameters data.activity.itempars
FIXED * fixed a bug in rasch.mirtlc with multiple starts in latent class models NOTE * allowed input of student weights and product-moment matrix in R2noharm FIXED * fixed a bug in calculation of degrees of freedom in R2noharm function FIXED * fixed a bug in linear transformation for person parameters in linking.haberman ADDED * included eigenvalue decomposition function 'eigenvalues.sirt'
DATA * included datasets: 'data.noharmExC', 'data.mcdonald', 'data.reck'
DATA * included data conversion function 'data.wide2long'
DATA * renamed 'data.noharm' into 'data.noharm18'
NOTE * corrected legends and labels in plot.isop function ADDED * included function rowKSmallest2.sirt FIXED * fixed a bug in mcmc.3pno.testlet (thanks to Chun-Hsien Wu) -> removed parametrization 3 (bifactor model) because it produced untrustworthy results NOTE * included gregmisc in imports (thanks to Jean-Paul Fox) NOTE * disabled parametrization 2 in mcmc.3pno.testlet -> now only the original model from the Glas paper (parametrization 1) is estimated NOTE * included argument for item parameter fixings in pairwise conditional likelihood function rasch.pairwise.itemcluster NOTE * included ETS classification labels of uniform DIF in function dif.logistic.regression NOTE * included the option of not submitting any item strata to stratified.cronbach.alpha and an example for using the psych package ADDED * included function 'md.pattern.sirt' for analyzing response patterns NOTE * included crossprod function in several sirt functions ADDED * included monotone nonparametric estimation in 'rasch.mml2' as the default in case of irtmodel='npirt' and npformula=NULL NOTE * speeded irtmodel='npirt' in case of providing npformula NOTE * included output of summary of testlet effects in mcmc.3pno.testlet ADDED * included the functional unidimensional model (Ip et al. 2013) ADDED * included the linking method of Haberman (2009) NOTE * included plot method for rasch.mml2 function NOTE * removed rasch.testlet.glmer function
ADDED * included estimation of ISOP and ADISOP models FIXED * fixed a bug in pbivnorm2 for negative correlations NOTE * included argument increment.factor in smirt to prevent nonconvergence (thanks to Balal Izanloo) ADDED * included rowwise maximum and minimum and rowwise cumulative sum of a matrix based on Rcpp code ADDED * included matrix functions rowIntervalIndex.sirt (useful for simulations) and rowKSmallest.sirt (useful for sorting matrix entries, e.g. for predictive mean matching in mice) FIXED * fixed a bug in mcmc.2pno.ml (thanks to Ronny Scherer) ADDED * included matrix row and column wise monotone regression function monoreg.rowwise (adapted from fdrtool package) ADDED * included graded response model estimation in isop.poly function ADDED * included plot method for ISOP models (isop.dich and isop.poly)
DATA * included example data.big5 EXAMP * included an example for DETECT for polytomous data
ADDED * included the estimation of the 2PL model via MCMC (mcmc.2pno) NOTE * included a function for estimating Rhat and MAP from MCMC output (mcmc.list.descriptives) ADDED * included a function for writing coda files (mcmclist2coda) NOTE * renamed jacknife.rasch.jml into rasch.jml.jackknife1 ADDED * included mcmc.2pnoh for estimating a hierarchical IRT model ADDED * included mcmc.3pno.testlet for estimating IRT testlet and bifactor models for dichotomous data ADDED * included mcmc.2pno.ml function for estimating a multilevel IRT model for dichotomous and normally distributed data ADDED * included R2NOHARM functions into this package for running NOHARM from within R NOTE * modelfit.sirt now allows objects of class smirt NOTE * speeded calculation of modelfit.sirt NOTE * speeded rasch.mml2, prob.guttman, rasch.mirtlc, rm.facets, rm.hrm
NOTE * speeded rasch.pml2 a bit FIXED * fixed a bug in smirt function (thanks to Izanloo Balal) ADDED * included function tetrachoric2 faster calculation of the tetrachoric correlation matrix ADDED * included pbivnorm2 for calculating the bivariate normal integral NOTE * included rasch.pml3, a bit faster version than rasch.pml2 FIXED * eliminated warning statements in smirt and fixed a bug in estimation of a and b parameters in smirt (thanks to Izanloo Balal)
NOTE * included a Rcpp function for faster calculations in rm.hrm (now there is a dependency of Rcpp and RcppArmadillo) ADDED * included smirt function for estimating the multdimensional noncompensatory and compensatory models ADDED * included variational approximation for the Rasch model ADDED * included the (Rasch) grade of membership model
ADDED * included function for computing eigenvalues and inverses of many matrices ADDED * included facets models with item/rater intercepts and slopes ADDED * included initial version of the hierarchical rater model based on signal detection theory (DeCarlo et al., 2011) ADDED * included function for simulating and estimating the Dirichlet distribution ADDED * included grade of membership model with joint maximum likelihood estimation -> gom.jml
NOTE * included standard errors in rasch.mml2 ADDED * included marginal true score reliability for binary item response data (Dimitrov, 2003; thanks to Ulrich Schroeders)
ADDED * included classification accuracy in the Rasch model ADDED * implemented the multidimensional latent class Rasch model (Bartolucci, 2007) NOTE * cleaning of help files, included several datasets and examples NOTE * included log-linear smoothing of the theta distribution in rasch.mirtlc ADDED * included 2PL estimation in rasch.mirtlc ADDED * included global fit statistics in modelfit.sirt NOTE * included log-linear smoothing of the ability distribution in rasch.mml2 NOTE * accelerated rasch.pml function and renamed it into rasch.pml2 NOTE * allowed for linear constraints for residual correlations in rasch.pml2 NOTE * accelerated latent.regression.em.raschtype and plausible value routines
NOTE * restructured rasch.mml2 function to get a bit more compact code NOTE * do not export rasch.mml and rasch.copula anymore via namespace ADDED * included nonparametric IRT function estimation in rasch.mml2 -> argument irtmodel='npirt' (Rossi, Wang & Ramsay, 2002) ADDED * included multidimensional models in rasch.mml2 NOTE * added corrected AIC (AICc) in rasch.mml2 and rasch.copula2 NOTE * corrected mean and SD estimation in rasch.mml2 FIXED * fixed a bug in the calculation of the number of estimated parameters in rasch.copula2 FIXED * corrected rasch.mml2 function in case of fixed weights of trait distribution FIXED * corrected a bug in wle.rasch ADDED * included probabilistic Guttman model ADDED * added analytical bias correction method for JML estimation of the Rasch model NOTE * accelerated PMML estimation algorithm ADDED * added 2PL estimation in rasch.pml
FIXED * fixed bad numerical behavior in rasch.copula2 for Cook-Johnson and Frank copula models NOTE * improved speed in yen.q3 and extended output NOTE * speed improvement in rasch.mml and rasch.copula2 NOTE * included version rasch.mml2 with substantial speed improvements NOTE * speed improvements for rasch.jml function
NOTE * changed algorithms behavior of rasch.copula2 (parameter increments depend on number of iterations) FIXED * corrected deviance calculation in rasch.copula
NOTE * removed functions for running ConQuest from this package ADDED * included Cook-Johnson and Frank copula in rasch.copula2 function ADDED * included examples for wle.rasch.jackknife FIXED * removed a minor printing bug in rasch.copula FIXED * fixed a bug in jackknife.rasch.jml NOTE * included item parameter constraints in jackknife.rasch.jml FIXED * corrected handling of extreme responses in rasch.prox
ADDED * included estimation of residual error correlations in rasch.pml
NOTE * restructured rasch.testlet.glmer function
DATA * included data set reading data.read
NOTE * set a different default in rasch.mml (estimation of trait SD seemed incorrect)
NOTE * rasch.testlet.glmer: output EAPs and EAP reliability
NOTE * included loglog and cloglog link functions in rasch.testlet.glmer
ADDED * included jackknife estimation of the joint maximum likelihood
NOTE * included a speeded version of rasch.copula and labelled it
as rasch.copula2 -> The old version rasch.copula remains in the package.
NOTE * output posterior distribition, EAP estimates and
EAP reliability in rasch.copula
ADDED * introduced probit and logit version of rasch.pml: pairwise marginal likelihood estimation FIXED * removed some bugs (or inefficiencies) in numerical differentiation in rasch.mml and rasch.copula NOTE * renamed sim.rasch.copula into sim.rasch.dep ADDED * included function rasch.pairwise.itemcluster for the conditional pairwise estimation method for locally dependent items
NOTE * package structure of sirtr is created ADDED * included simulation function for Rasch copula model ('sim.rasch.copula')
Version 0.21 -- 2012-06-23 Version 0.20 -- 2012-06-18 Version 0.19 -- 2012-04-26 Version 0.18 -- 2012-03-29 Version 0.17 -- 2012-02-22 Version 0.16 -- 2012-02-17 Version 0.15 -- 2011-12-19 Version 0.14 -- 2011-10-06 Version 0.13 -- 2010-08-26 Version 0.12 -- 2010-01-13 Version 0.11 -- 2010-01-12 Version 0.10 -- 2010-01-02 Version 0.9 -- 2009-11-01 Version 0.8 -- 2009-03-08 Version 0.7 -- 2009-03-02 Version 0.6 -- 2009-12-02 Version 0.5 -- 2008-11-12 Version 0.4 -- 2008-08-25 Version 0.3 -- 2008-05-06 Version 0.2 -- 2008-01-29 Version 0.1 -- 2007-12-22 Version 0.0 -- 2007-12-18