Latent Trait Models under IRT

Analysis of multivariate dichotomous and polytomous data using latent trait models under the Item Response Theory approach. It includes the Rasch, the Two-Parameter Logistic, the Birnbaum's Three-Parameter, the Graded Response, and the Generalized Partial Credit Models.


News

Changes in version: ltm_1.0-0

o small bug fixes.

==============================

Changes in version: ltm_0.9-9

o use of globalVariables() in source code.

o small bug fixes.

==============================

Changes in version: ltm_0.9-7

o changed deprecated sd() for matrices.

==============================

Changes in version: ltm_0.9-6

o minor bug fixes.

==============================

Changes in version: ltm_0.9-5

o a bug was corrected in fitted.gpcm(), fitted.grm(), fitted.ltm(), fitted.rasch(), and fitted.tpm(), i.e., fitted values are no longer rounded.

==============================

Changes in version: ltm_0.9-4

o rasch() has a new logical argument 'Hessian' which if set to FALSE, the Hessian matrix is not calculated. This could be useful in simulations.

==============================

Changes in version: ltm_0.9-3

o functions "ltm()", "rasch()" and "tpm()" no longer abbreviate the names of the variables.

==============================

Changes in version: ltm_0.9-2

o "ltm()" has been upgraded such that it can fit the Quotient model.

o a bug in "item.fit()" has been corrected.

o "factor.scores()" has two extra logical arguments: 'prior' which if TRUE (the default) the prior normal distribution is taken into account in the calculation of the ability estimates; 'return.MIvalues' which if TRUE it return the estimates abilities and their variance-covariance matrix for each one of the multiple imputations.

==============================

Changes in version: ltm_0.9-1

o the coef method for objects of class "grm" and "gpcm" now returns the estimated coefficients without rounding.

o a bug in the print method for objects of class "itemFit" and "persFit" that was no printing correctly in Linux and Mac systems, has been corrected.

o a bug in the print method for objects of class "rcor.test", has been corrected.

==============================

Changes in version: ltm_0.9-0

o there is a new function gpcm() (along with methods for the supporting generic functions) for fitting the generalized partial credit model.

o the new function GoF.gpcm() computes the Pearson chi-squared goodness-of-fit measure and also produces a p-value either using the asymptotice chi-squared distribution or a paremetric Bootstrap approach.

o the anova method for class 'gpcm' includes a simulate.p.value argument, which if is TRUE it computes an estimate to the p-value of the likelihood ratio test using paremetric Bootstrap.

o a bug was fixed in the fitted method for classes "grm", "ltm", "rasch" and "tpm" that did not work when the resp.patterns argument was a data.frame with factor columns.

o a method has been written for the simulate generic for objects of class "gpcm" and "grm".

==============================

Changes in version: ltm_0.8-9

o the calculation of the observed frequencies has changed in the presence of NA; side-effect: residuals() does not produce valid results for patterns with NAs. This will be rectified in a future release.

o plot.fscores() now alos works for patterns with missing data.

==============================

Changes in version: ltm_0.8-8

o a wrong warning message tha appeared after calling grm() in a data set with missing responses has be rectified.

o a bug was fixed in plot.descript() that did not work when the first argument of descript() was a matrix (a not a data.frame).

==============================

Changes in version: ltm_0.8-6

o changes in e-mail addresses.

==============================

Changes in version: ltm_0.8-5

o a bug was fixed in factor.scores() when the response patterns contain missing data.

==============================

Changes in version: ltm_0.8-4

o a bug was fixed in the print.tpm() method when argument IRT.param was FALSE.

==============================

Changes in version: ltm_0.8-3

o a bug was fixed in the application of the Delta method for computing the standard errors for 'grm' objects.

o factor.scores() no longer returns the observed and expected frequencies of the response patterns when missing data are present in the response data.

o some additions in .Rd files.

==============================

Changes in version: ltm_0.8-2

o a bug was fixed in "cronbach.alpha()" for the case both arguments 'standardized' and 'na.rm' are TRUE.

o "rcor.test()" has two new arguments, namely 'p.adjust' and 'p.adjust.method' that correct the reported p-values for multiple testing.

o a bug was corrected in the "factor.scores()" method when "testEquatingData()" has been used.

==============================

Changes in version: ltm_0.8-1

o the new function "cronbach.alpha()" computes Cronbach's alpha for given data-sets. The function can also compute a Bootstrap confidence interval for alpha.

o a "plot()" method has been added for 'fscores' objects. It plots a Kernel Density Estimation of the distribution of the factor scores (i.e., person parameters). Provides also the option to include in the plot the item difficulty parameters (similar to the Item Person Maps).

o "factor.scores()" have now the extra option (specified via the 'method' argument) to compute expected a posteriori scores.

o the "descript()" function also returns Cronbach's alpha estimates, and biserial correlation with the total score by excluding each time the item under consideration.

==============================

Changes in version: ltm_0.8-0

o "person.fit()" and "item.fit()" functions have been included that compute p-values for person- and item-fit statistics for IRT models for dichotomous data. The 'simulate.p.value' argument enables the computation of p-values based on a parametric Bootstrap procedure.

o "unidimTest()" function has been included that can be used to check the unidimensionality assumption for IRT models for dichotomous data using Modified Parallel Analysis.

o "testEquatingData()" function has been included for preparing data-sets for test equating by common items and common subjects.

o "grm()" works now with the available cases when incomplete response patterns (i.e., in the presence of NAs) are considered.

o better algorithms, for Missing At Random missing values, have been writen for "grm()", "ltm()", "rasch()" and "tpm()".

o "factor.scores()" and "fitted()" methods for classes 'grm', 'ltm', 'rasch', and 'tpm' allow for NAs in the 'resp.patterns' argument.

o a "residuals()" method has been added for 'grm', 'ltm', 'rasch', and 'tpm' objects that computed Pearson-type residuals.

o the "fitted()" method now allows also for the computation of marginal and conditional probabilities; this feature is controled by the 'type' argument.

o "rmvlogis()" has the an extra option to simulate under the U(-3.5, 3.5) distribution for the latent variable.

o "rmvlogis()" has the extra argument 'z.vals' that can be used to specify the values of the latent variable (ability) to be used in the simulation.

o the default value for the 'max.guessing' argument of "tpm()" has been set to one.

o package ltm requires now two extra packages, namely package mvtnorm and package polycor.

o a bug was fixed in "plot.ltm()" to allow for different colours for the Item Characteristic Surfaces.

o a bug was fixed in the "print()" method for 'ltm' objects that outputed the parameter estimates under the IRT parameterization, even when argument 'IRT.parm' was set to FALSE.

o a bug was fixed in "tpm()" that was not passing user-defined starting values correctly.

o an "ltm-package" Rd file has been included.

o additions and corrections in Rd files.

==============================

Changes in version: ltm_0.7-1

o a bug was fixed in "grm()".

o the "plot()" method for all IRT models has an extra logical argument 'plot' that controls whether the plot will be produced or not. If 'plot = FALSE', then the values used to create the plot are returned.

o the "plot()" methods for 'descript' objects has an extra argument 'items' specifying which items to plot.

o a better "print()" method for 'rcor.test' objects has been added.

o package 'gtools' is no longer required.

==============================

Changes in version: ltm_0.7-0

o function "biserial.cor()" has been included that computes the point biserial correlation between a dichotomous and a continuous variable.

o in the "plot()" method for 'grm' objects, the option for Operation Characteristic Curves has been added.

o "rmvlogis()" has a new argument 'IRT' specifying whether the parameters are under the IRT parameterization.

o "rmvlogis()" has extra available distributions for the latent variable.

o a "plot()" method has been added for 'descript' objects, producing the plot of Total Score versus Proportion of Correct for each item.

==============================

Changes in version: ltm_0.6-1

o starting values for "ltm()" are now by default deterministic; for random starting values you may use 'start.val = "random"'.

o "anova.rasch()" now only computes likelihood ratio test between nested models. The parametric Bootstrap goodness-of-fit test in now implemented in function "GoF.rasch()".

o "information()" returns now a list of class 'information' for which a "print()" method has been added.

o 'Lsat' and 'Wirs' data.frames have been renamed to 'LSAT' and 'WIRS', respectively.

o "rmvlogis()" has an extra argument 'link' that specifies the link function.

o additions and corrections in Rd files.

==============================

Changes in version: ltm_0.6-0

o "tpm()" function is introduced for fitting Birnbaum's Three Parameter model.

o "grm()" modified to handle dichotmous items as well.

o "rmvlogis()" function can be used to simulate binary variates under dichotomous IRT models.

o "descript()" returns more output.

o the "plot()" method for classes 'grm', 'ltm', 'rasch', and 'tpm' contains the new argument 'z' that specifies the latent variable values used in the plots.

o a bug was fixed in "coef.rasch()" and "coef.ltm()".

o a demo for "ltm()" has been added.

==============================

Changes in version: ltm_0.5-1

o the Gauss-Hermite rule has been upgraded; functions "grm()", "ltm()" and "rasch()" may produce slightly different parameter estimates from the previous versions.

o some issues regarding missing values and singular designs have been resolved in "start.val.rasch()".

o a bug was fixed in "coef.grm()".

==============================

Changes in version: ltm_0.5-0

o the "plot()" method for 'grm', 'ltm' and 'rasch' objects is now more flexible containing a number of extra arguments.

o starting values for "rasch()" are now by default deterministic; for random starting values you may use 'start.val = "random"'.

o the "coef()" method for 'ltm' and 'rasch' objects has an extra argument 'order' that sorts the coefficients according to the difficulty parameters.

o corrected some typos in .Rd files.

==============================

Changes in version: ltm_0.4-1

o "information()" now has the argument 'items' that specifies for which items the information should be computed in the specified interval. By default, the test information is computed.

o a bug was fixed in "print.grm()".

o a bug was fixed in "plot.grm()".

==============================

Changes in version: ltm_0.4-0

o "grm()" function is introduced for fitting Graded Response models under IRT.

o "plot()" function has a new argument, 'type', that indicates whether Item Characteristic Curves or Item Information Curves should be plotted. If 'items = 0', the Test Information Curve is plotted. Some extra arguments have been also introduced, like 'lty' and 'col'.

o Both "grm()" and "rasch()" report parameter estimates under the IRT parameterization, by default. "ltm()" does the same for the two parameter logistic model.

o "margins()" is now a generic function with methods for 'grm', 'ltm' and 'rasch' objects.

o "factor.scores()" has a new argument 'resp.patterns' that provides specific response patterns for which factor scores should be computed. "factor.scores.ltm()" has now as default method 'EB' instead of 'Component'.

o "descript()" can now also handle data sets with polytomous responses.

o the function "mult.choice()" has been introduced that can convert multiple choice responses to 0-1 data; this can be used in "ltm()" and "rasch()".

o the function "information()" has been introduced that computes the area under the test information curve.

o the function "rcor.test()" has been introduced that computes and tests for the significance of sample correlation coefficients.

o the demo for "rasch()" has been updated.

o corrected some typos in .Rd files.

==============================

Changes in version: ltm_0.3-1

o "coef.rasch()" and "coef.ltm()" now obey to the 'IRT.param' argument.

o standard errors are now reported also for the case where the dicriminitation parameters are fixed.

o a demo for "rasch()" has been added.

==============================

Changes in version: ltm_0.3-0

o a new logical argument -- 'IRT.param' (default to FALSE) -- has been introduced in both "ltm()" and "rasch()" which if TRUE reports the coefficients under the usual IRT parameterization.

o "plot.ltm()" and "plot.rasch()" improved: (i) if the column names of the input data are not NULL then they are used to denote the items, (ii) printing of legend is now optional -- see argument 'legend', (iii) if 'legend = FALSE', 'text()' is used to add as labels to the existing plot either the column names of the input data or numbers denoting the items, (iv) there is an argument 'items' controlling which items are plotted.

o 'constraint' argument in "ltm()" accepts now more general fixed value constraints. For more info check help("ltm", package = "ltm").

o a "fitted()" method has been introduced for both classes 'ltm' and 'rasch'.

o a "logLik()" method has been introduced for both classes 'ltm' and 'rasch'.

o a "vcov()" method has been introduced for both classes 'ltm' and 'rasch'.

o now robust standard errors estimation is done inside "vcov.rasch()" and "vcov.ltm()".

o "coef.ltm()" and "coef.rasch" have a new logical argument 'prob' which if TRUE, then the probabilities of a positive response for the median individual are returned.

o "descript()" now has a print method -- 'print' argument no longer available.

==============================

Changes in version: ltm_0.2-1

o "rasch()" now contains an extra argument 'constraint' that permits incorporation of fixed-value constraints. For more info check help("rasch", package = "ltm").

o "anova.rasch()" now also performs LRTs between constrained and unconstrained Rasch models.

o a "vcov()" method has been introduced for "rasch()" objects.

o Edited *.Rd files to use \method{} for generics.

o corrected some typos in .Rd files.

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.

install.packages("ltm")

1.1-0 by Dimitris Rizopoulos, 2 months ago


https://github.com/drizopoulos/ltm


Browse source code at https://github.com/cran/ltm


Authors: Dimitris Rizopoulos <d.rizopoulos@erasmusmc.nl>


Documentation:   PDF Manual  


Task views: Multivariate Statistics, Psychometric Models and Methods


GPL (>= 2) license


Depends on MASS, msm, polycor


Imported by CTTShiny, IRTShiny, PsyControl, ShinyItemAnalysis.

Depended on by IMTest, PCMRS, PerFit, QVM, catSurv, difR, infutil, integIRTy, irtoys, kequate, pcaPA.

Suggested by catIrt, equateIRT, equateMultiple.


See at CRAN