Skip to contents

Changelog

Source: NEWS.md

mirtCAT 1.15

  • Added a customStop function support to design input to allow custom termination of the CAT session. This can be used instead of returning NA in the customNextItem approach (requested by @jian-an-chen and Johan Braeken)

  • Changed $terminated_sucessfully to $GUI_terminated_sucessfully to avoid confusion, and added terminated_early element to person objects to indicate whether the CAT termination criteria resulted in an early termination

mirtCAT 1.14

CRAN release: 2024-07-12

  • The shinyGUI list’s forced_choice global specification in mirtCAT() has now been deprecated in favor of including this information in the df object. This allows explicit selection of which items should be presented in a forced choice format (requested by Levent on the mirt-package forum)

mirtCAT 1.12

CRAN release: 2021-12-08

  • Added design information to include the sequential probability ratio test for classification purposes. Currently limited to unidimensional models only

  • Fix Fisher information bug in unidimensional graded response model applications (reported by Felix Fischer)

  • New argument df$HTMLOptions added to indicate whether the Options.# inputs should be evaluated using HTML() first and passed to the suitable constructor functions (requested by John Manacup)

  • df$Timer is now printer for each respective item in the GUI, and if time runs out then the item is either replaced with an NA response or marked as “incorrect” if one or more Answer elements were present (requested by Morten Nordmo)

mirtCAT 1.10

CRAN release: 2020-07-10

  • Time remaining now reported in real-time, where now the last ten minutes of a GUI session prints the minutes and seconds remaining

  • Re-work the updateDesign() function to more naturally move through the CAT session manually. Previous response and selected item arguments now replaced with new_response and new_item, respectively, and example documentation updated to reflect this change

  • ordered and unordered constraints no longer stop in the middle of testlet set, thereby ignoring other stoping criteria (suggested by Christoph König)

  • choiceNames and choiceValues inputs required additional unnaming (reported by Ryan Kelley)

  • Up to three attempts are allowed when users are required to provide a password (requested by Aron Fi)

  • createShinyGUI() now contains a host_server logical argument to indicate whether the GUI is hosted on a remote server or locally. This triggers whether shiny::stopApp() should be executed upon completion (on servers this should be executed to avoid stopping the app when multiple users are active)

  • previously provided sessionName argument now removed in favour of handling this feature in the back-end. This helped patch a bug related to multiple instances of the GUI on a single server (reported by Daniel Sommerhoff)

  • thetas.start now accepts a matrix input for simulations with different latent trait starting values. Note that this also will affect the start_item criteria as well since the optimal first item will incorporate the new starting trait location (requested by Niek Frans)

mirtCAT 1.9.3

CRAN release: 2019-06-28

  • Added a element to person objects and results return from GUI version of mirtCAT() called clientData, which pertains to the character vector information from shiny’s session$clientData component (requested by Felix Fischer and Daniel Sommerhoff)

  • Allow multiple CAT session to be run during the same R session. This is primarily an update for users interested in hosting mirtCAT on a single server where multiple users can be processed at the same time

  • Fix design_elements global dependency between objects (reported by Felix Fisher)

mirtCAT 1.8

CRAN release: 2018-10-22

  • default wrapper to Questions element in df input now uses shiny::HTML rather than the previous shiny::p. This allows for HTML code to more naturally be included, as well as the use of shiny tag constructor functions when used in concert with as.character(). This deprecates the previous df$StemExpressions format

  • list inputs for choiceValues and choiceNames can accept named input elements that match the associated rownames in the df input to avoid the use of placeholders

  • The final page splash screen now removes the Next button to avoid confusion (suggested by Daniel Sommerhoff). Applications must be explicitly closed now by the user, and consequently the stopApp logical to the shinyGUI list has been deprecated

  • Fixed the random item selection exposure control, which was selection suboptimal items instead (reported by @machow on Github)

  • When Type input is text or textArea and forced_choice = TRUE then blocks of text input will also now be required. Otherwise, a message will be printed to the GUI

mirtCAT 1.7

CRAN release: 2018-03-11

  • message added in red whenever selection conditions not met (e.g., when using forced choice, rating scale options, or new mastery input)

  • added an optioanl Mastery input colum to df to prevent the GUI from continuing until the item is answered correctly (not useful for CATs, but supported for teaching interfaces)

  • added a rankselect GUI input to df$Type to create a list of rank-orded selection options corresponding to the input options

  • Fix bug when using sequentially scored test without forced choice (reported by Aron Fink)

mirtCAT 1.6

  • add Timer option to df intput to set time limits for select items

  • iterative algorithms now use the previous theta estimates as the starting values when locating the updated estimates

  • fix selection bug when using prior covariance matrix selection rules (reported by Johan Braeken)

  • computeCriteria() now respects the subset argument (reported by Johan Braeken)

mirtCAT 1.5

CRAN release: 2017-05-04

  • new choiceNames and choiceValues inputs to shinyGUI list to allow HTML stems for radio buttons and checkboxes (e.g., see the new arguments in help(radioButtons))

  • rotate x-axis 90 degrees in plot() generic by default

  • start_item can now be a vector to change the starting item for each off-line response pattern input in Monte Carlo simulations

  • fixed off-line starting item when set to 'random', which previously set each participant’s start item the same (reported by Okan Bulut)

  • internal updating function for person class to avoid direct assignment in customUpdateThetas() function

mirtCAT 1.4

CRAN release: 2017-03-02

  • warning message and slot included in GUI results when session terminated unexpectedly

  • fix min items combination when using a preCAT list input (total min items is now the sum of both stages)

  • fix scoping bug when hosting GUI on server (reported by Murat Doğan ŞAHİN)

  • when a tab is closed early in the GUI the session now calls stopApp()

  • new customUpdateThetas function input to design to create a map of how the latent trait estimates should be updated throughout the session. To better expose the internals, the previous Update.thetas() function associated with a person class is now available as a slot in the design internal object

  • added subset and info_mats to computeCriteria() to compute the desired criteria on item subsets only and to return the information matrices used to compute criteria such as Drule, Trule, etc (requested by Johan Braeken)

mirtCAT 1.3

CRAN release: 2017-02-01

  • removed support for a list input for df object in place of a more flexible format which supports the arbitrary creation of user-defined item formats through the new customTypes argument. This is to be used in conjunction with the df$Type input to pair the respective inputs

  • added time_before_answer to shinyGUI list to include a require number of seconds to wait before a valid/omitted response can be accepted

  • add textAreaInput support to GUI for include text-boxes as possible inputs

  • df$StemExpression is now instead a logical vector used to indicate whether the questions definition should be evaluated in R first

  • added AnswerFun input to allow for user-defined functions for each respective item to determine whether the answer provided is correct or incorrect. Mainly useful for text-based items (suggested by Anna Mikolajetz)

  • added a theme input to shinyGUI to support predefined themes from the “shinythemes” package

mirtCAT 1.2

CRAN release: 2016-11-16

  • progress logical added to mirtCAT() to print the progress of Monte Carlo simulations with the pbapply package

  • Shadow testing and optimal test assembly designs supported by adding constr_fun input to mirtCAT() and objective input to findNextItem(). See the findNextItem() documentation for details and examples

  • computeCriteria() function added to return all the criteria values (e.g., maximum information) associated with each respective item

  • population-level Theta terms tracked internally for Monte Carlo simulations because generate_pattern() now stores the respective population values as a silent attribute. Will now be displayed in summary/print/plot outputs, where applicable

  • allow the internal design object to be modified through the customNextItem() function

  • added an extract.mirtCAT() function to better document and extract the components of the person, design, and test objects in customNextItem(). Allows for safer and easier customization of item selection maps

  • new test_properties and person_properties inputs to design element to allow for prior information about the test/persons (e.g., selecting particular items based on prior demographic information). Intended for use with a customNextItem() function definition

mirtCAT 1.1

CRAN release: 2016-09-09

  • added StemExpression to df input to support arbitrary R expressions for stems to be rendered with standard R code. Allows for inputs such as audio, video, etc, as well through the use of a suitable tags$FUN(...)

  • new password input to shinyGUI list to allow login and password gates. Useful when hosting CAT on a server with potentially sensitive items

  • default shiny stem wrapper function for df when it is a data.frame is now exposed via the stem_default_format input to shinyGUI

  • MathJax is now explicitly supported in the item stems and response options. As before, users can provide a manual list explicitly defining the shiny functions; however, if a data.frame object is used then the stems/responses will be automatically wrapped within a suitable shiny::withMathJax() function to render the output

  • mental preparation screen now disabled for non-scored tests when the mo object is omitted (i.e., equivalent to setting shinyGUI = list(begin_message = ""))

  • 'select' Type input to df now uses a blank option as the default to avoid clicking through accidentally. Also works better with the forced_choice default

mirtCAT 1.0

CRAN release: 2016-07-26

  • released for JSS publication

mirtCAT 0.9

CRAN release: 2016-06-24

  • fix MLWI/MPWI selection criteria

  • findNextItem() is allowed to return all items given the selection criteria

Changes in mirtCAT 0.8

CRAN release: 2016-02-24

  • include constraints option for excluded items (useful for re-testing situations where item selection is from the same bank of items but previously answered items should not appear)

  • new 'none' presentation Type to allow empty material with no responses (mainly for presenting reading passages or other testlet-based items). Useful when used in conjunction with the constraints input from the design list so that these items have predictable dependent pairings (or can be used with the customNextItem form as well for explicit dependencies)

  • shinyGUI input gained a ui element to allow users to completely customize the graphical UI with shiny code

  • added customNextItem function to allow a completely customized method for declaring the item selection scheme

  • support checkbox input type for checkboxGroupInput() function. Supplying one or more answers will score the items in a partial credit style

  • include missing input options in df input for shiny (e.g., width, placeholder, inline). Removed 'radio_inline' input because this can now be accomplished by passing Type = 'radio' and inline = TRUE

Changes in mirtCAT 0.6

CRAN release: 2015-10-04

  • added getPerson() function to assign a person object following the use of createShinyGUI

  • the raw_responses vector returned for each participant is now a character vector instead of an integer

  • added slider Type input for sliderInput rating scale questions

  • resume_file removed, now if an associated temp_file is detected it will be used to resume the GUI session

  • new begin_message input to shinyGUI to change message prior to starting the CAT

  • changed lastpage element in shinyGUI input from a list to a function to allow messages containing information pertaining to the session

Changes in mirtCAT 0.5

CRAN release: 2015-05-09

  • fix bug when using SH exposure control when the pool runs out of items

  • add delta_thetas argument to design list to stop CAT if change between latent trait updates is less than this distance criteria

  • include more functions to allow easier web hosting in locations such as shinyapps.io. Namely, mirtCAT_preamble() and createShinyGUI()

  • graphical images (png, gif, and jpeg) no longer supported. Stems preferably should in HTML for best results and control

  • Stem paths can now point to and render HTML and markdown files in the GUI

  • df input can now be a list for including more fine tuned shiny inputs

  • check for the existence of graphical item stems, and allow relative or absolute paths

BUGFIX

  • fix stem location bug, and add better aspect ratio

  • the summary(mod, sort=FALSE) argument returned a NULL object instead of the unsorted response. This has been patched and is now tested for

  • when no df is supplied, check that the responses are in a plausible range (reported by Gerard Flens)

Changes in mirtCAT 0.4.2

CRAN release: 2015-01-25

  • forced_choice argument for shinyGUI input to state whether response for each item are required (for CATs, this should always be TRUE)

  • response in GUI now are blank by default

  • in preCAT list input change nitems to min_items and max_items for better control. Also include a response_variance logical to terminate the preCAT stage when variability in the the response options occurs (so that ML estimation becomes plausible)

  • include Sympson-Hetter method of item exposure control

  • major re-write of shiny inputs. Now the questions, answers, and options are all supplied through a data.frame object for better clarity, and all shiny inputs regarding the questions, image locations, etc, are specified in this object

  • mirt_object input renamed to mo for short

  • the local_pattern input now accepts a matrix of response patterns for running simulation designs. Returns a list of person objects instead of a single object

  • generate_pattern() now supports multiple Theta inputs specifically for simulation designs

  • start_item input now accepts character value inputs (in addition to explicit numeric inputs) corresponding to the criteria argument. This allows the first item to be selected using the (adaptive) selection criteria

  • added generate.mirt_object() function to build the required mirt_object input given population parameters

  • new cl argument in mirtCAT() for passing a parallel cluster object defined from the parallel package. Used to run simulation designs with parallel architecture for better speed

  • by default, the demographics page is no longer generated in the GUI

Changes in mirtCAT 0.3

CRAN release: 2014-12-15

  • item_answers can now be a list input, indicating that more than one correct answer is possible for a given item

  • allow the first page and demographics page to be skipped by passing empty list arguments

  • added ‘fixed’ method to keep latent trait estimates at fixed values (useful for preCAT)

  • Fisher information matrix added for remaining multidimensional models supported by mirt, including custom item types

  • add ‘Arule’ and ‘APrule’ for minimum trace criteria of asymptotic covariance matrix

Changes in mirtCAT 0.2

CRAN release: 2014-10-15

  • temporary files can now be saved while the GUI is running, and restored at a later time

  • more estimation options can be passed to fscores() via the … argument

  • sensitive objects are now removed from the package namespace when the mirtCAT() finishes unsuccessfully

  • categories are always returned with based 1 for first response in the GUI

  • add content balancing option

  • various bug fixes, and update documentation

  • new findNextItem() input for users to locate the next item to administer (likely for custom CAT interfaces that do not use the Shiny package). Can be updated with the updateDesign() function

  • moved ‘Next’ button in the web interface to the left panel box so that it will always remain in the same location

  • support CSS customization

  • switch multidimensional selection criteria to use analytical expressions rather than numerical. Several multivariate Fisher information matrix computation currently not supported analytically, but will be steadily added.

  • add classification capabilities to design list input