************************************************** * * * Changes and Developments in the catR Package * * * ************************************************** ------------------------------------- - Changes in catR 3.17 (2022-06-22) - ------------------------------------- MAJOR UPDATES o Package maintainer changed from David Magis to Cheng Hua o Test the new compling environment using R studio MINOR UPDATES o Some typos were fixed ------------------------------------- - Changes in catR 3.16 (2018-06-29) - ------------------------------------- MAJOR UPDATES o The 'randomCAT()' function holds a new argument called 'a.stratified' that allows a-stratified sampling for dichotomous IRT and GRM, MGRM and GPCM polytomous models. This argument sets either the number of strata or the number of items per strata (as a vector of integer values). The item bank is split into strata by increasing values of the discrimination (a) parameters and content balancing is then applied with equiprobable strata. o The new function 'aStratified()' was included to return a vector of stratum membership that is used for a-stratified content balancing. o The 'print.cat()' function was updated accordingly. MINOR UPDATES o Some typos were fixed. ------------------------------------- - Changes in catR 3.15 (2018-06-12) - ------------------------------------- MAJOR UPDATES o The 'randomCAT()' function holds a new argument called 'min.length' that specifies the minimum length of the CAT (i.e., minimum number of items to administer). Checks that minimum length is fulfiled in addition to the 'checkStopRule()' function. Default value is zero so no lower limit is set for the test length. MINOR UPDATES o Some typos were fixed. ------------------------------------- - Changes in catR 3.14 (2018-03-24) - ------------------------------------- MAJOR UPDATES o The 'thetaEst(') function now supports the robust estimator of ability. Two weight functions are allowed, Huber and Tukey weights. Tuning constants can be set appropriately too. Note that the robust estimator is currently available only for dichotomous IRT models. o The 'semTheta()' function holds new formulas for computing the ASE of the BM and WL estimators. These formulas arise from Magis (2016) and are made available through the new 'sem.type' argument. Usual formulas are still available by default and only dichotomous IRT models are concerned. o A new function called 'fullDist()' is available. It derives the full distribution of an ability estimator for a given set of items, a set of ability estimates and several parameters to defined the estimation method. This function is routinely used to compute exact standard errors. o The 'semTheta()' function allows now the computation of exact standard errors for dichotomous IRT models. Such exact values are computed according to Magis (2014). This is activated by setting the new argument 'sem.exact' to TRUE. o The 'test' list of the 'randomCAT()' function holds now the 'se.ase' argument that specifies the maximum number of items to be used for computing exact SE, before switching to ASEs for sake of computational effort. o Functions 'testList()', 'randomCAT()' and 'simulateRespondents()' were modified accordingly. MINOR UPDATES o Main author's affiliation and email address was updated. o The list of contributors of some functions was updated to include Lianne Ippel (University of Liege, Belgium). o Some typos were fixed. ------------------------------------- - Changes in catR 3.13 (2017-08-31) - ------------------------------------- MAJOR UPDATES o A bug was fixed in the 'Ii()' function. For polytomous IRT models, the second derivatives of the information function were incorretcly computed due to a typo. This bug was only affecting the computation of SE values for the BM estimator with Jeffreys prior only, all other estimators and prior distributions were unaffected. Many thanks to Xu Kun (University of Hong Kong; xujacob.hk@gmail.com) for noticing this bug. o A new item bank was introduced to this version. It is called 'cat_pav' and contains 96 polytomous items calibrated under the GPCM. Moreover it holds item names as row names of the matrix. Further info is available from the 'cat_pav' help file. Many thanks to Victor D.O. Santos (Iowa State University, Ames, IA) for sharing his item bank. o Many help files using polytomous IRT models were updated: the generated PCM item bank was replaced by the 'cat_pav' GPCM item bank. o Item names can now be supplied as the row names of the item bank. Otherwise no item name is considered. o Functions 'startItems()' and 'nextItem()' now return the additional argument, called 'names' with the former function and 'name' with the latter function. It contains either the names of the selected item(s) whenever these are available, or 'NULL' value otherwise. o Function 'randomCAT()' has an additional output argument called 'itemNames'. It contains, whenever available, the names of the items that were selected during the CAT. Moreover, the 'print.cat()' function was updated so that when item names are available, they are displayed in the summary table of the output (instead of their numbered position in the item bank). MINOR UPDATES o Some help files were corrected as they were still containing details related to former implementations e.g., 'randomCAT()' function). o Some typos were fixed. ------------------------------------- - Changes in catR 3.12 (2017-01-07) - ------------------------------------- MAJOR UPDATES o The function 'startItems()' was updated to include the arguments 'cbControl' and 'cbGroup' from function 'randomCAT()' for instance. This is to allow random selection of the first item(s) by approximately matching the theoretical proportions of items per subgroup. Moreover, the argument 'random.cb' was added to fix the random seed when selecting the subgroups of items from which to sample the items. o The function 'testList()' was updated accordingly. o In the function 'randomCAT()' the 'start' list holds two new arguments: 'cb.control', a logical value to indicate whether content balancing control should be done with starting items, and 'random.cb' argument from 'startItems()' function. o The function 'simulateRespondents()' was updated accordingly. MINOR UPDATES o The DESCRIPTION file was updated to meet CRAN policies. o The CITATION file was updated for publication in Journal of Statistical Software. o The help file references were updated. o The writing of DOI reference numbers was updated. o Some typos were fixed. ------------------------------------- - Changes in catR 3.11 (2016-11-16) - ------------------------------------- MAJOR UPDATES o The function 'simulateRespondents()' was updated to allow the 'start' and 'test' input lists being exactly the same as in 'randomCAT()' function. Several recent updates of 'randomCAT()' were not passed to 'simulateRespondents()' through its input lists, which is now fixed. MINOR UPDATES o The DESCRIPTION file was updated to incorporate more details about the scope of the package. o Some typos were fixed. ------------------------------------- - Changes in catR 3.10 (2016-08-26) - ------------------------------------- MAJOR UPDATES o The code of 'startItems()' function was updated to fix a bug that was occuring in very specific cases. The selection of the optimal starting items (based on the rank of selection criteria) was written in such a way that it was sometimes not selecting any item (because of ranks larger than the selection thresholds in case of ties), leading to an error. Many thanks to Kristin M. Morrison (Georgia Institute of Technology, Atlanta, GA; kmorrison3@gatech.edu) for noticing this issue. o The code of 'randomCAT()' function ws updated to fix a bug occuring when the 'nrItems' element of 'start' list was fixed to zero. In this case, the 'checkStopRule()' function was leading to a logical error. Many thanks to Andrew Kyngdon (BOSTES and Graduate School of Education, University of Western Australia; andrew.kyngdon@bostes.nsw.edu.au) for noticing this issue. MINOR UPDATES o Some typos were fixed. ------------------------------------ - Changes in catR 3.9 (2016-07-15) - ------------------------------------ MAJOR UPDATES o Fixing the seed when selecting item(s) with the 'randomesque' method was added in the startItems() and nexItem() functions. The argument 'random.seed' may take a numeric value to fix this random selection, or set to NULL to avoid fixing the seed. Many thanks to William Lorie (Pearson) for noticing this issue. o Functions testList() and randomCAT() were updated according to the modifications above. MINOR UPDATES o The 'print.catResult' from simulateRespondents() function was updated to display the appropriate dichotomous IRT model that is used (formely only '4PL' was returned). o Some typos were fixed. ------------------------------------ - Changes in catR 3.8 (2016-06-10) - ------------------------------------ MAJOR UPDATES o The setting of stopping rules was updated. It is now possible to expand the 'stop$rule' vector with more than one stopping rule accronym, so that the CAT stops once at least one of the rules is satisfied. The 'stop$thr' must also be expanded accordingly with the corresponding stopping thresholds. o A new testing function called 'checkStopRule()' was introduced. It is used internally to test whether at least one of the stopping rules was satisfied. o Given the two updaters above, the main argument 'maxItems' of the randomCAT()' and 'simulateRespondents()'functions was removed, and the 'testList' function was updated accordingly. o A bug was fixed in the 'randomCAT()' function as a consequence of the modification in the previous update. When several items were selected at the start, a perfect response pattern was generated, and the stepsize adjustment approch was chosen, then an error was occurring in computing the ability estimate. MINOR UPDATES o The 'print.cat()' function had an issue in displaying the output appropriately in case of constant pattern at the start of the CAT. This has been fixed. o The help file of 'startItems()' was updated to correct a former description of the arguments. o Some typos were fixed. ------------------------------------ - Changes in catR 3.7 (2016-05-09) - ------------------------------------ MAJOR UPDATES o The functions 'thetaEst()' and 'semTheta()' now allow missing responses in the 'x' pattern. They must be coded as NA and are discarded from the estimation process (most useful for linear testing use). o The function 'genPattern()' can now generate several patterns at the same time, by providing several 'th' values as a vector. The output is then a matrix of item responses instead of a single vector. o The function 'startItems()' was modified to simplify the setting of the initial theta values to select the first items. Now the 'theta' argument can hold a vector of values, and one item is picked up for each element of 'theta'. The argument 'halfRange' was consequently withdrawn from this function as it became useless. o The function 'startItems()' holds a new argument called 'randomesque' and acts similarly to the 'randomesque' argument of the 'nextItem()' function. That is, if it takes values larger than one, the 'randomesque' most optimal items are selectd for eahc theta value, and one of them is picked up randomly as starting item. Default value is one so no such randomesque selection is done. MINOR UPDATES o In case of constant pattern and 'constantPatt' argument different from 'NULL', the'plot.cat()' with 'ci' argument set to TRUE was leading to a bug (infinite confidence intervals were obtained). This has been fixed internally into 'plot.cat()'. o Some typos were fixed. ------------------------------------ - Changes in catR 3.6 (2016-01-02) - ------------------------------------ MAJOR UPDATES o The function 'Pi' was updated to correct a major bug with some polytomous IRT models. More precisely, the 'Pi' function held an error for models 'PCM', 'GPCM' and 'RSM', leading to incorrect values for the derivatioves of response probabilities. This was subsequently impacting the values of functions 'I', 'Pi', 'thetaEst' and 'semTheta', among others. Many thanks to Norman Verhelest (Eurometrics; (norman.verhelst@gmail.com) for noticing this issue. ------------------------------------ - Changes in catR 3.5 (2015-09-27) - ------------------------------------ MAJOR UPDATES o The function 'thetaEst' holds three new arguments, 'constantPatt', 'current.th' and 'bRange'. These arguments are required to make apropriate theta estimation adjustments when the response pattern is constant. Currently implemented only for dichotomous IRT models. o The function 'semTheta' holds now the newly added argument 'constantPatt'and the functions 'testList', 'randomCAT' and 'simulateRespondents' were updated accordingly. o Two new item selection methods were added: the Global Discrimination Index (GDI) and the posterior global discrimination index (GDIP). Corresponding fucntion 'GDI' was added and functions 'nextItem' and 'testList' were updated. MINOR UPDATES o A bug was fixed about random seed in the 'start' list. When 'seed' was provided to the 'start' list it affected the whole CAT, leading to identical patterns (i.e. erasing the 'genSeed' argument). This has been fixed (i.e. the 'seed' argument of 'start' list only determines the selection of the first items). o The function 'genDichoMatrix()' allows now the $props elements of the 'cbControl' argument not to sum to one. This argument is internally normalized to one. o Most of the examples in the help files are now exempt from the \dontrun{} section. Only time-consuming examples are still not run. o Some typos were fixed. ------------------------------------ - Changes in catR 3.4 (2014-11-26) - ------------------------------------ MAJOR UPDATES o The 'simulateRespondents' was modified internally to fix a bug occuring when not all arguments of the 'start', 'test', 'stop' and 'final' lists were passed to the function. o The argument 'nrItems' of function 'genPolyMatrix' was replaced by 'items' to match with function 'genDichoMatrix'. Moreover, by-default values were asigned to arguments 'items' (100) and 'nrCat' (3) of 'genPolyMatrix'. o The 'genPattern' function has now a 'seed' argument to possibly fix the random seed for generating item responses. By default it is set to NULL so no random seed is determined. o The 'randomCAT' and 'simulateRespondents' functions have now a 'genSeed' argument to possibly fix the random generation of item responses. By default 'genSeed' is set to NULL so no random seed is determined (and two successive runs of 'randomCAT' and 'simulateRespondents" will usually lead to different response patterns). o If post-hoc simulations are run with 'randomCAT' function and the 'trueTheta' argument is not provided, then it is returned as NA in the output list. MINOR UPDATES o The help file of function 'genPolyMatrix' was rewritten to match with help file of function 'genDichoMatrix'. o Some typos were fixed. ------------------------------------ - Changes in catR 3.3 (2014-11-26) - ------------------------------------ MAJOR UPDATES o When the item bank holds several items with the same parameters, the 'MFI' next selection rule was often selecting the same item several times. This bug has now been fixed. Many thanks to Karthikeyan Madathil (kmadathil@gmail.com) for noticing this issue. MINOR UPDATES o The reference to Chang and Ying (1996) and Barrada, Olea, Ponsoda, and Abad (2010) were corrected in the 'nextItem' help file. o The DESCRIPTION file was modified to match CRAN rules. ------------------------------------ - Changes in catR 3.2 (2014-09-10) - ------------------------------------ MAJOR UPDATES o A bug was fixed in the 'randomCAT' function. When more items had to be administered than the length of the item bank, an error was occuring. The maxLength variable within 'randomCAT' was updated accordingly, as well as the 'print.cat' function. o A bug was fixed in the 'eapEst' function. With polytomous IRT models the 'model' argument was not passed properly. Many thanks to Przemyslaw Lis (The Psychometrics Center, University of Cambridge, UK) for noticing this issue. o An error was corrected in the 'MEI' function. The expected infomation was not correctly computed with dichotomous IRT models (i.e. ability level 'th0' was used instead of 'th1' in the observed information function). Many thanks to Dirk Pelt (Erasmus University Rotterdam, The Netherlands) for noticing this issue. o A bug was fixed in the 'genPolyMatrix' function. When different number of response categories were generated for the items, the 'seed' argument was not operating properly. MINOR UPDATES o Some typos were fixed. ------------------------------------ - Changes in catR 3.1 (2014-06-04) - ------------------------------------ MAJOR UPDATES o The scaling constant D was not properly specified in 'startItems' and 'nextItem' functions. This has been fixed. o A bug was occuring when 'nrItems' was set to zero in the 'startItems' function. In this case, the methods 'MLWI', 'MPWI', 'KL' and 'KLP' of 'nextItem' were not working properly. Moreover, the 'plot.cat' function was not working when 'nrItems' was fixed to zero. Finally, the output of 'simulateRespondents' was also affected by this problem and was corrected. o In addition, the scaling constant D was embedded to all polytomous IRT models (but the NRM). All subsequent functions (such as 'Pi', 'Ii' and so on) have been updated. Many thanks to Mike Powell (mike.powell@learninglogistics.com) for suggesting this update. MINOR UPDATES o The DOI numbers were added to the references (whenever available). o Some typos were fixed. ------------------------------------ - Changes in catR 3.0 (2014-04-25) - ------------------------------------ MAJOR UPDATES o Functions 'startItems', 'nextItem' and 'randomCAT' have been updated thanks to Juan Ramon Barrada (Universidad Zaragoza, Spain), which is now listed as co-author of the package. o Function 'nextItem' was also updated thanks to Guido Corradi (Universidad Zaragoza, Spain). o 'catR' now supports the Graded Response Model (GRM; Samejima, 1969), the Modified Graded Response Model (MGRM; Muraki, 1990), the Partical Credit Model (PCM; Masters, 1982), the Generalized Partial Credit Model (GPCM; Muraki, 1992), the Rating Scale Model (RSM; Andrich, 1978) and the Nominal Response Model (NRM; Bock, 1972). o Functions 'Pi', 'Ii', 'OIi', 'thetaEst', 'semTheta', 'eapEst', 'eapSem', 'startItems', 'EPV', 'MEI', 'MWI', 'nextItem' and 'randomCAT'allow now the use of polytomous IRT models. o The function 'createItemBank' has been withdrawn from this version. The item bank must now be provided directly as the matrix of item parameters (plus the group membership for content balancing, optionally). o Five new item selection rules have been added: Kullback- Leibler (KL), posterior Kullback-Leibler (KLP), the so- called thOpt (optimal theta) method, progressive method and proportional method. They are available for both dichotomous and polytomous IRT models (except thOpt). o The 'KL' function was created to support the computation of Kullback-Leibler information for next item selection. o Several CATs can be run in one step by using the new 'simulateRespondents' function. In addition, summary statistics related to all generated tests are provided by this function. Corresponding plots can also be drawn and saved in a similar way as with 'randomCAT'. o The dependency to the 'sfsmisc' package was withdrawn. To do so, an internal function for numerical integration, called 'integrate.catR', was written and works similarly to the 'integrate.xy' function from package 'sfsmisc' (which was considered so far). In other words, the function 'integrate.xy' was replaced by 'integrate.catR' in 'eapEst', 'eapSem' and 'MWI'. o The functions 'genDichoMatrix' and 'genPolyMatrix' were created to partly compensate the withdrawal of 'createItemBank'. Those functions allow for easy generation of item banks under either the dichotomous IRT models (1PL, 2PL, 3PL and 4PL) or the polytomous IRT models that are supported by 'catR' (GRM, MGRM, PCM, GPCM, RSM and NRM). The output format is identical to the one required in order to run CATs with dichotomous and polytomous item banks. Random allocation of items into subgroups (for content balancing purposes) is also possible. o The function 'breakBank' was created to separate the item banks in two parts, the matrix of item parameters and the vector of group membership for items. This function works with both dichotomous and polytomous IRT banks. Its creation resulted from the withdrawal of the 'createItemBank' function. o The function 'Ji' was created to compute the J function (related to WL estimation) under dichotomous and polytomous models. This function was previously embedded into 'thetaEst' for dichotomous models, but is now written independently for both dichotomous and polytomous models. o The function 'test.cbList' has been updated. The second argument does not need to be the full item bank anymore but just a vector of group membership (such as the fifth column of the 'tcals' data set only). This is to remove the connection with the 'createItemBank' function (that has been withdrawn from this update) and to allow its use with polytomous models. o The functions 'EPV', 'MEI', 'MWI', 'startItems' and 'nextItem' have been updated. The argument 'itemBank' argument does not need to be the full item bank anymore but just a matrix with four columns with item parameters (such as the first four columns of the 'tcals' data set only). This is to remove the connection with the 'createItemBank' function (that has been withdrawn from this update). o The default value of the 'startSelect' argument of the 'startItems' function has been changed from "bOpt" to "MFI" for compatibility with the 'startItems.poly' function. o The function 'genPattern' was created to generate responses patterns under dichotomous and polytomous IRT models. This is helpful mostly for examples but can also be considered internally in 'randomCAT'. o The function 'startItems' can now select only a fraction of the item bank to select the first items. This is controled by the new 'nAvailable' argument, holding 0's and 1's for items that are respectively not available and available. By default, all items are available. o The random seed for the selection of the first item(s) in 'startItems' can be set to NA in order to remove random seed selection of these items. o The 'nrItems' argument of 'startItems' can now be set to 0 to avoid early selection of starting items. Only NULL values are then returned. o The function 'randomCAT' can now return also ability level estimates and standard errors for the first starting items. This is done by fixing the argument 'allTheta' to TRUE (default value is FALSE). If so, provisional ability estimates and standard errors are displayed within parentheses in the output of 'randomCAT' (i.e. when using 'print.cat()' function). MINOR UPDATES o Currently the formula for the SE of the WL method in the 'semTheta' function is the same as that for the ML method, as it happened that the formula (in versions of catR prior to 3.0) was sometiems leading to mathematical problems. This formula will be updated once it will be accurately corrected. o The 'startItems' function was slightly modified. When items are randomly selected, the random sampling of these items was rewritten using the 'sample' function. o In 'nextItem function, the Urry's procedure to select the next item was renamed as 'bOpt' method (instead of 'Urry'), for homogenbeity with the similar argument of the 'startItems' function. o The help file of the 'MEI' function was updated as an error was present in the description of both the observed and the expected item information function. Their implementation, however, was correctly written. o When the first items are selected randomly or fixed by the user, the 'startSelect' element of the output list returns now NA instead of the by-default value of the argument 'startSelect'. o To save the output of 'randomCAT', the 'output' argument contains now three components, the file path, the file name, and the type of file (in this order). By default, the output is stored in working directory, as a csv file with name 'name.csv'. Other type is text file. o The order of the components of 'save.options' argument of 'plot.cat' has been modified to match those of the 'output' argument of 'randomCAT'.. o Some typos were fixed. o Author's affiliation was updated. ------------------------------------ - Changes in catR 2.6 (2013-12-03) - ------------------------------------ MINOR UPDATE o An error was debugged in the 'nextItem' function. With 'MLWI', 'MPWI', 'MEI' and 'MEPV' the set of parameters of previously administered items was not correctly passed to the corresponding functions in case of content balancing (that is, vector 'OUT' was used instead of vector 'out'). This was leading to warning messages and incorrectly computed values of the criteria. Many thanks to Irshad Mujawar (Arcadix Infotech Private Limited, Bangalore, India) who noticed this problem. ------------------------------------ - Changes in catR 2.5 (2013-02-07) - ------------------------------------ MINOR UPDATES o An error was debugged in the 'MWI' function. The selection of the information function values to compute the MLWI or MPWI was wrong, so that the same item was always selected, independently of the current response pattern (for the same set of administered items). Many thanks to Koken Ozaki (Institute of Statistical Mathematics, Tokyo, Japan) who noticed this problem. o Some typos were fixed. ------------------------------------ - Changes in catR 2.4 (2013-01-17) - ------------------------------------ MINOR UPDATES o An error was debugged in the 'createItemBank' function. When creating difficulty parameters with the 'unif' distribution, the selection of specified parameters was leading to an error. Many thanks to Przemek Lis and Michel Kosinski (Cambridge University, UK) who noticed this problem. o Some typos were fixed. ------------------------------------ - Changes in catR 2.3 (2012-05-09) - ------------------------------------ MINOR UPDATES o The references of several help files have been updated with respect to the formal publication of the related JSS paper about catR. o The CITATION file has been updated accordingly. ------------------------------------ - Changes in catR 2.2 (2012-03-15) - ------------------------------------ MINOR UPDATES o The references of several help files have been updated. o A basic NAMESPACE file has been added for compatibility with R versions 2.14.0 and later. ------------------------------------ - Changes in catR 2.1 (2011-12-05) - ------------------------------------ MINOR UPDATES o It is now possible to save the output of 'randomCAT' into a text file, by specifying the new arguments 'save.output' and 'output' properly. Similarly, the plots obtained with the 'plot.cat' function can be saved as PDF or JPEG files by specifying properly the new arguments 'save.plot' and 'save.options'. See the corresponding help files for further details. o Some typos have been fixed in the help files. ------------------------------------ - Changes in catR 2.0 (2011-10-21) - ------------------------------------ MAJOR UPDATES o Several distributions are now available for item parameter generation of the 'createItemBank' function. Possible distributions are: normal and uniform for item difficulties; normal, lognormal and uniform distributions for item discriminations; uniform and Beta for both item lower and upper asymptotes. o Item exposure control is now feasible with the so-called 'randomesque' approach (Kingsbury and Zara, 1989). The function 'nextItem' and 'randomCAT' hold now the 'randomesque' argument that sets the number of optimal items to be selected first, among those the next administered item will be randomly chosen. o Content balancing can now be controlled by specifying subgroups of items in the 'createItemBank' function. The subgroups are coded into a fifth column (after item parameters) and they are stored in the item bank by setting the new argument 'cb' (for "content balancing") to TRUE. If this fifth column is missing, or if 'cb' is set to FALSE (default), nothing changes from the previous version of catR. See also functions 'nextItem' and 'randomCAT', which were modified for content balancing, as well as the new function 'test.cbList' for testing the format of the 'cbControl' list of 'nextItem'. o Item parameters can now be provided as data frames in the 'createItemBank' function. This also permits to include subgroups of items easily. o As a consequence of the previous major update, many help files were updated for comptability with the new function 'createItemBank' (i.e. 'tcals' is no longer needed to be converted into a matrix). o The 'tcals' data set was updated by adding a fifth column, holding the subgroups of items as defined by the test. See the help file of 'tcals' for further details. o As a consequence of the previous major update, many help files were updated for compatibility of the new 'tcals' data set (i.e. all examples requiring the matrix of item parameters were slightly modified to remove the 'Group' variable). o The 'print.cat' function was updated to display output results of randomesque method for item expoisure control as well as for content balancing (whenever required). MINOR UPDATES o When the 'maxItems' criterion of 'randomCAT' is smaller than the 'stop$thr' argument, an error message is returned. Previously it contained a mistake, claiming that 'maxItems' is larger than 'stop$thr'. This has been fixed. o Some typos have been fixed in the help file of 'tcals' data set and 'nextItem' functin. ------------------------------------ - Changes in catR 1.6 (2011-04-11) - ------------------------------------ MINOR UPDATES o The 'Pi' functin has been slightly modified to avoid a technical bug: when probabilities Pi equal exactly zero or one, the ability estimates cannot be obtained with 'thetaEst' (the boundary conditions cannot be tested because of NaN values). Many thanks to Pan Tong (University of Texas MD Anderson Cancer Center, USA) who noticed this problem. ------------------------------------ - Changes in catR 1.5 (2010-10-28) - ------------------------------------ MINOR UPDATES o The help files of the 'thetaEst' and 'semTheta' methods have been updated as some information about the EAP estimation method was missing. ------------------------------------ - Changes in catR 1.4 (2010-10-17) - ------------------------------------ MINOR UPDATES o The method referred to Owen's rule for selecting items in the starting step and in the adaptative process, was renamed Urry's rule given that this simple method was suggested first by Urry (1970) and that Owen's Bayes procedure differs from this one. o Consequently to the point above, 'Owen' optional value of the 'itemSelect' argument of the 'test' list was replaced by 'Urry'. o In addition, the 'nextItem', 'testList', randomCAT' and 'print.cat' have been updated by the replacement of 'Owen' by 'Urry'. o A minor problem was fixed in the code of the 'randomCAT' function: the output list was not returning all the required elements (in particular, the value of the 'itemSelect' argument). o A minor issue was fixed in the 'print.cat' function: the output was not displayed correctly when only two items are selected at the initial step. ------------------------------------ - Changes in catR 1.3 (2010-10-08) - ------------------------------------ MINOR UPDATES o Several typos have been fixed. o The 'randomCAT' help file has been updated as some information was incorrect or incomplete. ------------------------------------ - Changes in catR 1.2 (2010-09-07) - ------------------------------------ MAJOR UPDATES o Several procedures for next item selection have been added. First, Maximum Likelihood Weighted Information (MLWI) and Maximum Posterior Weighted Information (MPWI) are now available. The command 'MWI' has been developed for that purpose. o Second, Maximum Expected Information (MEI) has been implemented for next item selection. The function 'MEI' was developed for computing MEI values. o The MEI criterion required the creation of the 'OIi' command for computing observed information function. o The Maximum Expected Posterior Variance (MEPV) is also available. The command 'EPV' computes expected posterior variances and is used within the 'nextItem' function. o Finally, the completely random selection of the next item is also possible right now. o The functions 'eapEst' and 'eapSem' have been modified to speed up their computation. The approximation of the integrals is now made by the 'integrate.xy' function from the package 'sfsmisc'. o The function 'startItems' has been modified. First, the argument 'bw' has been replaced by 'halfRange' and codes now for the half range of the sequence and not the full range (this was misleading). Second, the items can be selected either according to their difficulty level or their information function. This is specified by the 'startSelect' argument. o Consequently, the functions 'nextItem', 'testList', 'randomCAT' and 'print.cat' have been updated. MINOR UPDATES o The argument 'eapPar' from functions 'eapEst' and 'eapSem' has been renamed to 'parInt' for consistency with the 'MWI' and 'nextItem' functions. It better refers to the parameters for numerical integration and not only for the EAP method, but also for MLWI and MPWI computation. o Several typos have been fixed in the help manual. o The 'plot.cat' function has been modified. Now, the confidence intervals do not appear by default ('ci=FALSE' in the code). Moreover, confidence intervals are displayed with dotted traits, while the true ability level is drawn with a solid horizontal line. Also, any classification level can be specified and drawn by a dashed horizontal trait. ------------------------------------ - Changes in catR 1.1 (2010-06-15) - ------------------------------------ MAJOR UPDATES o Owen's procedure has been added as a method for next item selection. o Consequently, the functions 'nextItem', 'testList', 'randomCAT' and 'print.cat' have been updated. MINOR UPDATES o Several typos have been fixed in the help manual.