Changes in version 1.6.0 Major updates - Ordinal IRT models were implemented including GPCM, PCM, RSM, and GRM. Comparison of nested models (GPCM, PCM vs. RSM) is available. Minor updates - Data upload logic improved. - Data not uploaded to global environment. Bug fixes - Fixed misplaced legend in the DD plot (in the interactive app). - plot_corr() no longer warns about unknown labels. - Reports work now even when Correlation structure is requested. - Item names with special characters should be now properly handled in PDF report. Changes in version 1.5.5 (2025-07-10) Bug fixes - Data upload was fixed. - Installation of the modules from the app now works even in R sessions without any CRAN mirror set. Changes in version 1.5.4 (2025-02-08) Major updates - R >= 4.1.0 is now required. - Modules installation is now provided in the application itself in the Modules section. This feature is only enabled when the application is run locally. Minor updates - ShinyItemAnalysis does no longer reexport %>% operator from magrittr package. The native pipe operator |> in used instead, internally. - The confidence level for Cronbach's alpha in the Reliability section is now stated. - If the interactive application is run locally, the visitors counter is no longer displayed. - Wording in PDF/HTML reports was edited. Bug fixes - Correlation test in Item analysis / Item criterion validity is now properly formatted. - Criterion type recognition in Item analysis / Item criterion validity is now fixed (criterion with less than 7 unique levels is considered as discrete variable). Changes in version 1.5.3 Minor updates - New methods are offered in Generalized logistic regression in the DIF/Fairness section of the app. {difNLR} version 1.5.0 is now required for the DIF analysis. Changes in version 1.5.2 (2024-11-27) Bug fixes - The starting values and estimation constraints structure in fit_blis() function were fixed to work with the new version of the mirt package (versions 1.43 onwards). - Missing parts of the Logistic regression note in "DIF/Fairness -- Method comparison" tab of the app were fixed. Changes in version 1.5.1 (2024-04-24) Major updates - run_app() function discovers for add-on modules available on SIA repository which are not installed yet, and offers their installation. Minor updates - There are now documented options and environment variables that are consulted by ShinyItemAnalysis package, see ?ShinyItemAnalysis_options. - In Data/Data tab, new section called Current data was added to explicitly track the current data used in the app and its source. Bug fixes - The number of iterations (EM cycles) for IRT models provided in the Settings is now considered in the app again. - The app won’t crash anymore when the user uploads nominal data without the key. Changes in version 1.5.0 (2023-05-08) Bug fixes - Calculation of itemfit p values was fixed in the Table of estimated parameters in IRT model/Dichotous models section of the app (Thanks to Samsul Hadi for reporting this issue). - Calculation of worst split-half coefficient was fixed to respond to updates in the psych package. Major updates - Additional modules can now be easily incorporated into the Shiny app! Shiny app discovers for installed modules and adds them in respective sections. Modules can be kept in external packages. See SIAmodules R package for default modules. - New dataset CLoSEread6 was added and it is also offered as a toy dataset in the Data/Data section of the app. - DIF analysis now offers combinations (simple and iterative) of purification and p-value adjustment in the DIF/Fairness section of the app. Minor updates - The plot_corr() function is now more verbose and handles errors in polychoric correlations estimation in a more user-friendly way; the partial matching of arguments is no longer supported - Data upload and treatment of missing values was clarified and slightly updated in Data/Data section of the app. - All data are now downloadable in Data/Data exploration section of the app. - Error messages caused by missing values were made more specific in DIF section of the app. - Summary table of total scores was complemented by number of complete cases and explanations in the Scores/Total scores and DIF/Observed scores section of the app. - Cutscore for hiding loadings was lowered to 0.2 in Validity/Factor analysis section of the app. - Interfactor correlations were added for oblique rotation and n_factors > 1 in the Validity/Factor analysis section of the app (Thanks to Stanislav Ježek for suggesting this feature). - HCIlong dataset now contains zscore variable. - Some of the HeightInventory dataset variables were renamed, new variable total was added. - Dependency on the psychometric package was removed. - Anxiety dataset is now offered as a toy dataset in the Data/Data section of the app. - Dataset CZmaturaS is newly offered as a toy dataset in the Data/Data section of the app. Changes in version 1.4.2 (2022-10-21) Bug fixes - Item criterion validity - Distractor plot was fixed in Item Analysis tab. Major updates - New datasets HCIlong, MSATB, GMAT, EPIA added - anxiety dataset was renamed to Anxiety, variables score and zscore were added - Variable total was added to datasets HCI, HCItest Minor updates - Parallel analysis plot has an improved legend in the Validity / Factor analysis tab Changes in version 1.4.1 (2022-04-15) Bug fixes - gDiscrim() function was fixed to assure indices are inside the defined bounds, also for small samples - plotDistractorAnalysis() function was fixed not to display double legends. Thanks to Scarlett Escudero for reporting this issue. - Item difficulty/discrimination plot in Tradition item analysis tab in the app was fixed to prevent error messages when Discrimination type is set to "none". - Displayed item parameters were fixed and subsequently extended in the app for the Bock model (see Major updates). - The training sections for GRM IRT model was fixed to display category plots also for case of k = 6. Major updates - The ItemAnalysis() has been refactored substantially; memory demands are lowered multiple times, as well as run times (most noticeable with large datasets). The changes include following: - The function now returns NAs for "gULI" when k, l, and u arguments are not specified, as there are no sensible defaults. Previous defaults resulted in exactly the same values that ULI already provided on its own. - The Index.rel.drop is not provided anymore, as it is not well defined (see Gulliksen, 1950, for more details on reliability index (Index.rel)). - The bin and add.bin arguments were removed, the analysis on binarized data can be conducted by simply providing the cutscore in the cutscore argument. - The criterion argument now defaults to NULL instead of "none", but behaves the same. - Deprecated arguments data, y are now defunct. - The DistractorAnalysis() and plotDistractorAnalysis() functions have been improved: argument items allows options "all" to apply for all items (default), or a vector of item names (column names of Data), or item identifiers (integers specifying the column number). - The Anxiety dataset was added. - Baseline Logit Slope-Intercept (BLIS) parametrization for Bock nominal response model is now available as a package function. Fitted model supports IRT reparametrization (to so-called BLIRT) using a standard coef() method as well. - The app training sections for NRM IRT model now includes parameters in parametrizations and identification restrictions assumed by Bock (1972), Thissen et al. (2011), as well as BLIS and BLIRT parametrizations accounting for the correct option. - The DIF section in the app was updated to work with the newest difNLR package version. Minor updates - gDiscrim() newly issues a warning when item(s) range is zero and when there is no difficulty difference in either group. - In gDiscrim(), the groups are newly defined using the quantile() function. In the case of ties, the respondents with the same total score are now in the same group, the groups sizes may thus differ, but are now exactly defined. The function may however provide different numerical results than in the previous version. - In gDiscrim(), you can now change how the dataset is split by passing arguments to findInterval(). - RStudio Addin was added to provide command-free app startup option, bindable to a keyboard shortcut. - The HCI data now includes total score variable. Changes in version 1.4.0 (2021-09-27) It includes version 1.3.8 and following further changes: Bug fixes - Math in downloadable tables is now completely correctly transformed to human-readable (non-LaTeX) format everywhere (thanks to Larry Nelson for reporting this issue). Major updates - The package dependencies that are used only in the interactive app per se were moved into Suggests, meaning you are now not obliged to install any of these dependencies if you plan to only use the package's "core" functions, such as DDplot() or ItemAnalysis(). - The startShinyItemAnalysis() function now checks for dependencies required for its proper run and offers to install these dependencies if any of them is not available. - New dataset HeightInventory was added. Minor updates - Manuals of R functions and data were updated Changes in version 1.3.8 Bug fixes - Better treatment of missing values in fa_parallel and in the Validity/Factor analysis section of the app, preventing crashes of the app (Thanks to @koenderks for reporting this as issue #460) - Math in downloadable tables is now partly correctly transformed to human-readable (non-LaTeX) format, including the IRT tab. Major updates - startShinyItemAnalysis() gained an alias run_app(). - New dataset AttitudesExpulsion was added. - DIF-C SIAmodule newly available from the DIF/Fairness section of the interactive app - IRR-restricted SIAmodule replaced the Reliability/IRR-restricted section of the interactive app - Math typesetting is now completely carried out by KaTeX, a lightweight and reliable library. By eliminating MathJax, the app now loads faster and no math is left unrendered. Minor updates - Sample R code was added for the Validity/Correlation structure section. - Manuals of R functions and data were updated - The app now uses trivial Cronbach alpha internal function, lowering the number of dependencies - www folder was cleaned up, .png replaced by .svg files improving the resolution of plots - README.md assets moved to .readme_assets folder - inst folder structure was simplified Changes in version 1.3.7 (2021-05-31) Bug fixes - Bug with reading ordinal data in the app was fixed. (Thanks to pukelevicius for reporting on GitHub) - Bug with uploading data with all correct/wrong answers in some items was fixed (Thanks to Koen Derks for reporting the issue on GitHub) Major updates - New datasets MSclinical, BFI2, and TestAnxietyCor were added, some datasets and manuals were updated. - Functions fa_parallel() was updated, smaller changes were made in some other functions - In the Validity section of the app, tab Factor analysis was updated. - In the IRT section of the app, tab for the Bock model was updated to offer parametrization closer to regression models. Minor updates - Sample R code was updated for the IRT section. Changes in version 1.3.6 (2021-03-20) Bug fixes - Function ItemAnalysis() updated to avoid error messages in case of datasets including 2 items only. (Thanks to Frederic Hilkenmeier for reporting this issue.) - Data upload in the app was fixed, as well as the warning messages related to data upload. - In the Reports section of the app the default name of the dataset was set to the name of the csv file. This fixed some of the bugs with the HTML report download. Major updates - New function fa_parallel() was added offering fast paralel analysis for deciding on number of factors in exploratory factor analysis. - New datasets CZmatura and CZmaturaS were added. - Some UI code of the shiny app was reorganized into separate sections. - In the Validity section of the app, a new tab Factor analysis was added offering exploratory factor analysis. - In the IRT models section of the app, the binary IRT models were collapsed into a single Dichotomous models tab. - In the IRT models section of the app, the Bock IRT model section was edited. - In the DIF/Fairness section of the app, the logistic and the polytomous regression models now offer both the IRT and the intercept-slope parametrization; the total score is newly offered as the matching criterion besides the standardized total score. - In all sections of the shiny app, the wording was edited. - The ggWrightMap() function was rewritten to exclude dependencies other than ggplot2. Minor updates - The sample R code snippets were moved to a separate directory. - In the Criterion validity and Item criterion validity, the Pearson correlation is newly used as the default option. - In the shiny app, the item criterion validity analysis was moved to the Item analysis section. - In the Item analysis section of the app, default discrimination was changed to RIT. - Softer deprecation with warning only was applied to functions. - In the shiny app, term Observed ability is now consistently used instead of (DIF/DDF) Matching variable. The changes done were especially in the Data, Regression models and the DIF/Fairness sections of the app, and in the plotDIFLogistic() function. - Section Summary of the app was renamed to Scores to beter reflect focus of this section. - In the Validity section of the shiny app, the dendrogram is newly enabled even if numclust == 0. - In the Item Analysis section of the app, the tab Distractors plot was renamed to Item response curves to better reflect that empirical response curves are also available for binary and ordinal items. Changes in version 1.3.5 (2021-02-03) Bug fixes - Plots in the DIF/About section of the app were fixed (uniform vs. nonuniform DDF is now correctly displayed). - Display of the discrimination index ULI was fixed for ordinal items. - Bug in display of Multinomial regression parameters in case of the IRT parametrization was fixed. - In the IRT training section, the calculation and display of item information curves was fixed. (Thanks to Hynek Cigler for reporting this issue.) - Calculation of minimal scores in the ItemAnalysis() function was fixed. - Display of ICC in IRT models was updated to respond to updates in the mirt package. - In the Regression section of the app, missplaced SE for intercepts and slopes were fixed. Major updates - A new section Reliability/Restricted range was added to the app offering interactive examples demonstrating issues with reliability estimates in restricted-range samples. - Continuous items are newly allowed, rater variable was added, data upload and warning messages were updated. - New function ICCrestricted() for computing ICC in restricted samples was added. - Functions gDiscrim(), DDplot() and ItemAnalysis() were updated, variable names were unified. - New function plot_corr() to compute and visualize an item correlation matrix was added and implemented within the ShinyItemAnalysis interactive app. - New datasets AIBS, NIH, HCIgrads, HCItestretest, HCIprepost and HCIdata were added. - startShinyItemAnalysis() function was fixed to finish job persisting in the backgroud after the app is closed. - Figures can now be downloaded at once in DDF section of the app. - Equations were unified in the Regression, IRT and DIF/Fairness sections of the app, with p being the index for person and i being the index for an item. - Interactive training section was added to the DIF/Fairness section, offering interpretation of group-specific models. - Figures in the Validity, Regression, IRT, and DIF/Fairness section are now displayed in plotly mode in the shiny app. - The IRT/Rasch section was restructured and it now offers both the intercept/slope as well as the IRT parametrization. (This is planned for other IRT subsections in future versions.) - The Regression/Multinomial section newly offers also the IRT parametrization. Minor updates - The intropage was updated, new News section was added. - Method ordering was updated in the DIF/Fairness section. - Toy data selection was updated. - Selected R code was updated. - Unnecessary dependencies were removed. - Summary tab was updated within the app, interactive help with formulae is newly offered. - Interpretation of IIC for selected theta were added in the IRT/training section. - styler package was applied on .R files, documentation was updated. - ggWright() plot was updated to more general description. - References were updated wintin the app. Changes in version 1.3.4 (2020-08-24) Bug fixes - Display of Cronbach's alpha in PDF/HTML reports was corrected. - Item slider bug in ordinal regression models in the app was fixed. - DIF matching criterion was synchronized for the Summary and Item subtabs for the Generalized logistic regression method in the DIF/Fairness section of the app. - Bug in display of table in the Regression/Model comparison section was fixed. Major updates - startShinyItemAnalysis() function was updated to allow starting the application as a background job in Rstudio (default setting). This permits the user to work and try sample R code examples in the console. - Downloadable tables are newly provided in shiny app in the DIF/Fairness sections for ordinal DIF. - DDplot is now displayed in plotly mode in the shiny app. - New function recode_nr() for recognizing and recoding not-reached items was introduced. - ItemAnalysis() function was updated with new columns for ratio of missed and unreached items, shortened column names and ordering. - Traditional item analysis table in the Item analysis section of the app has been expanded to include item reliability, item validity, and percentage of missing and unreached for each item. - DDplot() function now offers argument criterion. Without criterion specification, the difficulty-discrimination plot is provided. When criterion is specified, the difficulty-validity plot is provided instead, depicting item criterion validity. - Difficulty-validity plot is newly offered in the the Validity/Criterion_Validity/Items section. Minor updates - Documentation of DDplot() was updated. - All datatable outputs now have proper bootstrap style/theme. Changes in version 1.3.3 (2020-05-04) Bug fixes - plotDIFLogistic() now correctly plots matching criterion when item purification is applied. Major updates - Use of DIF matching criterion other than Total score is now enabled in shiny app in the DIF/Fairness sections. - Downloadable tables are newly provided in shiny app in the DIF/Fairness sections. - Iterations of purification process are now displayed in downloadable tables in the DIF/Fairness sections of shiny app. - DIF method comparison table is provided in the DIF/Fairness section of the shiny app. Minor updates - Help page for the ShinyItemAnalysis package was added. - NEWS file is now formatted using markdown. - On attach message was updated. - Wording, Figures and page structure were updated in shiny app in the DIF/Fairness sections, histogram of total scores now better compares the groups. Changes in version 1.3.2 (2020-01-28) It includes versions 1.3.1-1 - 1.3.1-8 and following changes: Bug fixes - Mismatch in specification of H0 and H1 was corrected in DIF/Fairness section and in the Report section. - plotDIFLogistic() now correctly plots matching criterion. - Logistic method in the DIF/Fairness section now correctly plots best fitting model when p.adjust.method is applied. - Logistic method in the DIF/Fairness section now excludes missing values. Major updates - New dataset LearningToLearn was added to the package. - New toy datasets LearningToLearn6 and LearningToLearn9 were added to the app. - Upload of the DIF matching criterion is now available in Data section of the app. - D-DIF index calculation was added into DIF/MH/Item to evaluate DIF effect size - OR calculation in DIF/MH was edited. - plotDIFLogistic() function was updated to accept class of Logistik as input. - plotDIFLogistic() now accepts character input for item argument. - Summary and Item inputs in DIF/Logistic are connected now. Model is now fitted only once. Minor updates - Citation and list of References were updated. - Authors' details were updated. - Due to changes in difNLR, function ddfORD()was renamed to difORD(). - About page was updated. - Error messages for DistractorAnalysis() and plotDistractorAnalysis() were unified. - Copyright year in footer is now responsive. - Description of toy datasets was updated. - styler package was used for all functions. Documentation was updated. - For plotDIFLogistic() it is now possible to turn off plotting empirical probabilities. This can be done using draw.empirical = FALSE. - Selected R code was updated in DIF/MH - Threshold for discrimination can be now specified in DDplot() function via argument thr. Default value is 0.2. - Warning and error messages were updated for plotDistractorAnalysis() function when key is missing. - Width of inputs was unified in DIF/Fairness section. - Wording was updated. - Some typos were fixed. - Text in reports was updated. - Some examples were set to run to allow for automatic checks of the package. Changes in version 1.3.1-8 Bug fixes - Typos in plotAdjacent(), plotCumulative() and plotMultinomial() were fixed. Changes in version 1.3.1-7 Major updates - Functions DistractorAnalysis() and plotDistractorAnalysis() were updated to not requiring key input. Changes in version 1.3.1-6 Bug fixes - Bug in key uploading was fixed. When nominal data was uploaded and then ordinal (without cut-score specified in file), key was kept from nominal data. - Bug in ordinal data was fixed. Recalculation of factors to numeric is done. - Some calculations with NA values were fixed. - Double file input "key" was unified. Minor updates - Item names for toy dataset Science were updated. - Basic summary table for ordinal data was updated. Changes in version 1.3.1-5 Bug fixes - Table for MH in PDF reports was fixed. - Header of sections were fixed in PDF reports. - IRT parameter estimates table was fixed in PDF reports. - Bug in plotMultinomial() for data without NAs was fixed. Major updates - DIF/DDF detection methods for ordinal data were added. DIF/Fairness section was updated. Minor updates - Tabs in Regression were renamed. - Warning when group vector is missing was added into Method comparison tab in DIF/Fairness section. - Method comparison tab for DIF/Fairness section was updated. Changes in version 1.3.1-4 Bug fixes - In DIF section, Summary table for total scores now displays correct number of respondents in groups (N). Changes in version 1.3.1-3 Bug fixes - Bug in Adjacent model in Regression tab was fixed. Matching criterion and parametrization inputs now have effect on model. - Bug in plotMultinomial() was fixed. Function now displays more than 12 response patterns as linetype. Minor updates - Range for x-axes in plotAdjacent() and plotCumulative() were updated to avoid removing data. - Some typos were fixed. Changes in version 1.3.1-2 Major updates - Cronbach's alpha with CI was added into reports. - Bonferroni correction for multiple comparison was added as an option for correction methods in Regression, DIF and Reports tabs. Minor updates - Warning for table for parameters in Bock's IRT model was modified. - Wright map now considers item names when option Keep item names. - Some wording was updated. - Some typos were fixed. Changes in version 1.3.1-1 Bug fixes - Minimal number of groups for gULI in DDplot was set to 2 in both Item analysis and Report tabs. Major updates - Mantel-Haenszel test was added into reports. Minor updates - Error for DIF/Logistic was customized. - Text in reports in DIF section was updated. - Tables in DIF section now includes both p-values and adjusted p-values when correction method is applied. - Warning for fancyhdr latex package was solved. Changes in version 1.3.1 (2019-06-26) It includes versions 1.3.0-1 - 1.3.0-18 and following changes: Minor updates - In ItemAnalysis() function, documentation for item reliability and item validity indices has been added (Reported by Lot Fonteyne). Changes in version 1.3.0-18 Bug fixes - Bug in plotDistractorAnalysis() was fixed. X-ticks are now displayed even with discrete = T argument. Minor updates - Inputs in Validity/Correlation overlaped when window was narrow. Fixed. - Generate and Download buttons overlaped in Reports when window was narrow. Fixed. Changes in version 1.3.0-17 Bug fixes - DistractorAnalysis() and plotDistractorAnalysis() now handle discrete matching and groups specified via argument cut.points. - Distractor plots in Validity/Items tab now properly handle discrete criterion and the number of groups is correctly descreased when criterion does not have unique cuts. - Bug in Raju in DIF/Fairness tab was fixed. Plots for items are now displayed for dataBov dataset. Major updates - Checking of unique cuts for distractor plots was added into Reports. Warning is displayed when cuts are not unique and slider is updated. Minor updates - Some typos were fixed. Changes in version 1.3.0-16 Bug fixes - Bug in Bock's model for binary items was fixed in reports. - Histogram when some levels of total scores are missing were fixed in Summary and DIF tab. Major updates - Legend was added into IRT plots when downloaded. - IRT models now respect ordering of items. Minor updates - Some typos in IRT models tab were fixed. Changes in version 1.3.0-15 Bug fixes - DIF summary table bug was fixed. Table is disaplayed when some DIF methods cannot be fitted. - DDF parameter tables were fixed for data BOV and Czech Trivia. - With difNLR package version 1.2.3 plots for DDF are displayed even for binary data. Plots were also fixed for data Bov. - PDF Reports now work locally and on shinyapps.io. - Dendrogram now properly replaces item names when Keep item names is not selected in Data upload section. Minor updates - Authors' and contributors' details were updated. - Function plotMultinomial() was updated to return smooth curves for any matching criterion. Changes in version 1.3.0-14 Bug fixes - Bug in Bock's nominal IRT model was fixed. ICCs are now properly displayed for all binary data. Error for non-displayable table of parameters was further improved. Tables for items are displayed even in case that table for all items is not. - DIF items are now correctly written in reports. - Corrplot now properly replaces item names when Keep item names is not selected in Data upload section. - In DIF/Fairness section, histograms of Total scores for groups are now properly displayed also for ordinal items. Major updates - Legends for multinomial and distractor plots were improved. Minor updates - Authors' and contributors' details were updated. Changes in version 1.3.0-13 Bug fixes - Bug in Bock's nominal IRT model was fixed. ICCs are now properly displayed for Czech Trivia dataset. Error for non-displayable table of parameters was improved. - Bug in DistractorAnalysis() was partially fixed, need to be revised. Major updates - Reports: - long tables have now header on each new page - legend is now visible for IRT plots even when dataset consists of more than 50 items - Matching criterion in plotMultinomial() was relaxed to decrease size of reports. Changes in version 1.3.0-12 Bug fixes - Reports are now rendered even in case that multinomial model cannot be fitted for some items. - With option Keep item names, names of items are properly displayed in Report for DIF detection methods. - Model comparison table returns now also partial results when some models cannot be fitted. - Bug in plotMultinomial() was fixed. In case of only two categories observed, plot is displayed properly now. - Bug in DistractorAnalysis() was fixed. In case of possibly discrete matching, matching is cutted based on its factors. Major updates - Uploaded data can be now unloaded without restarting app. - Distractor plot for validity analysis was updated to allow for lowering the number of categories in cases when cuts are not uniquely defined. Application was updated accordingly. Minor updates - Errors and warnings were customized. - In all dichotomous models the number of EM cycles can be increased in Settings. - Seed was set for validity boxplot for reproducibility of jitter points. - All static DOI links were updated. (Thanks to Katrin Leinweber) - Some typos were fixed in reports. Changes in version 1.3.0-11 Bug fixes - Option for displaying session info in reports now functions properly. - Option to display average score in DDplot in reports fixed. Major updates - New function plotCumulative() which plots category probabilities functions estimated by \code{multinom()} from nnet package was added. It is now used in application. - New function plotAdjacent() which plots category probabilities functions estimated by \code{multinom()} from nnet package was added. It is now used in application. - New function plotMultinomial() which plots category probabilities functions estimated by \code{multinom()} from nnet package was added. It is now used in application. Minor updates - Selected R codes for cumulative, adjacent and multinomial models were updated. - Option for matching criterion (total vs Z-score) in multinomial model was added. - Equations in DIF section are now center-aligned. Changes in version 1.3.0-10 Major updates - Alerts for analysis where data with missing values are omitted were added. - DDplot() now can displayed average item score as estimate of difficulty for ordinal data. See argument average.score. Minor updates - Correlation structure now use pairwise complete observations for calculation of correlation. Changes in version 1.3.0-9 Major updates - DistractorAnalysis() and plotDistractorAnalysis() functions were updated to allow for lowering the number of categories in cases when cuts are not uniquely defined. Application was updated accordingly. Minor updates - Tables and equations of polytomous models in Regression section are now center-aligned. Changes in version 1.3.0-8 Bug fixes - With option Keep missing values unfunctioning analyses were fixed (missing data removed, warning message added) Major updates - Multinomial regression plot is now displayed using smooth function Minor updates - With option Keep missing values warning message is now displayed for those analyses for which missing values are omitted. - For ordinal data, Cronbach's alpha is now calculated from original ordinal (non-binarized) data both in Reliability and in Item analysis section. Changes in version 1.3.0-7 Bug fixes - Busy indicator now properly displays above input tabs. - Bug when using matching in DistractorAnalysis() was fixed. Major updates - New ordinal toy dataset SCIENCE was added. Minor updates - Each tab is now open at the top of the page. Changes in version 1.3.0-6 Major updates - Adjacent logistic regression model was implemented for ordinal data. - Both adjacent and cummulative regression models allow for selection of matching criterion (total score, standardized total score) and parametrization (classical slope/intercept and IRT parametrization). Changes in version 1.3.0-5 Bug fixes - SD in Regression section was changed to SE. - Keep missing values button is now functioning and it is applied for ordinal and binary data. Major updates - Cumulative logistic regression model was implemented for ordinal data. - Delta method is now applied for calculation of SE in IRT parametrization in cumulative logistic regression. Changes in version 1.3.0-4 Bug fixes - Bug in slider in histogram of total scores was fixed. Range is now correctly displayed when NA values are present. - Fixing versions. Major updates - Math mode is now stable in tables. (Thanks to Stéphane Laurent) - UI for polytomous models in Regression section was added. - Traditional item analysis table is done via ItemAnalysis() function. It now includes also average score. Minor updates - startShinyItemAnalysis() automatically launches default web browser. - Sample code and wording was updated for traditional item analysis table in app and also in reports. Changes in version 1.3.0-3 Minor updates - ggWrightMap() now can display user-specified names of items (Reported by Matthew G R Courtney) - plotDIFLogistic() now can display user-specified names of reference and focal group. - Some typos were fixed. Changes in version 1.3.0-2 Major updates - Analysis for ordinal items has been added. This includes use of ordinal data in sections Summary (histogram of Total scores, success rate), Reliability, Validity (correlations, predictive validity), Changes in version 1.3.0-1 Bug fixes - Equations for 3PL and 4PL regression models were fixed. Major updates - Math symbols are now displayed using MathJax. - Exercises were added in IRT section / Polytomous Training. - DIF section now includes method comparison. Minor updates - Labeling of correction methods in DIF section was improved. - Ability parameters are now available also for Bock's model. Changes in version 1.3.0 (2019-02-25) It includes versions 1.2.9-1 - 1.2.9-3 and following changes: Minor updates - Reference to a new paper on ShinyItemAnalysis in The R journal was added. Changes in version 1.2.9-3 Bug fixes - Histograms for total scores now handle cut-scores out of range of observed values - Bug in reports when using delta plot with purification was fixed. - Cut-scores for histograms are properly displayed in reports. - Year in reports was updated. Major updates - Sections About, References and Settings are displayed only with icon and are right-aligned. - Margins and padding were updated. - t-test was added for comparison of total scores between groups into DIF section and into reports. - Histograms for total scores are now displayed in plotly mode. Minor updates - Some math mode was fixed in IRT section. - Some typos were fixed. - Selected R code was updated in DIF/Total scores. - Animate function for sliders was slowed down. Changes in version 1.2.9-2 Bug fixes - Generalized ULI index is now correctly displayed in traditional item analysis table in reports. - Item names are now correctly displayed in reports. - Item names are now correctly displayed in DDF section. - Function plotDIFLogistic() now handles user-specified matching criterion. Minor updates - Legend of DDplot was modified. Changes in version 1.2.9-1 Minor updates - Wording in PDF/HTML reports was edited. Changes in version 1.2.9 (2018-12-07) Bug fixes - IRT/Training/Dichotomous models - plots are correctly plotted. - Download buttons in IRT/4PL works now. Major updates - TIF were converted into plotly. - Section Validity/FA is not accessible for now. Minor updates - Selected R code for IRT/Bock's model was updated. - Wording in IRT models was updated. - Wording in Reports was updated. - Legend for plots of IRT models was modified in Reports. - Ability estimates summary table is now round in Reports. - Some typos were fixed in Reports. - Colours of IRT summary plots correspond to model for Items. Changes in version 1.2.8-7 Major updates - IRT section of shiny app was updated and now provides new and more shiny figures in plotly - PDF and HTML reports were updated and now allow - different types of correlation in corrplots - different types of discrimination indices in DDplots - summary table for ability estimates in IRT section Changes in version 1.2.8-6 Bug fixes - Dendrogram is now correctly rendered. Major updates - Function ItemAnalysis() was added. - IRT section now allows download of ability estimates Minor updates - Help pages for functions were updated. Changes in version 1.2.8-5 Bug fixes - Key for ordinal data for binarization is now correctly used. Values greater or equal to value of the key are set to 1. - When NULL data is uploaded, default dataset GMAT is in use. Major updates - Scree plot was moved into the new tab Factor analysis. - Dendrogram was added into Correlation structure tab. - Option for global cut-score for ordinal data was added. - R package ggdendro is now used. - Removal of items with only 0s or 1s now works. - Removal of observation with NA group now works. Minor updates - Text in Data section was updated. - List of references was updated. - Selected R code in Correlation structure tab was updated. Changes in version 1.2.8-4 Bug fixes - Bug in IRT models/Training/Dichotomous models text of theta in plotly was fixed. Major updates - Reorganization of data upload and renaming data variables. test_answers()> nominal() correct_answers()> binary() scored_test()> total_score() test_key()> key() DIF_groups()> group() criterion_variable -> criterion() And creating new variables: ordinal() ordinal data maximal() maximal values for ordinal data minimal() minimal values for ordinal data z_score() standardized total score, no need to use scale(total_score()) in each function Minor updates - Updating text in Data section. - Editing boxes for data upload in Data section. Changes in version 1.2.8-3 Bug fixes - Bug in Regression/Nonlinear 4P IRT Z (rounding in interpretation) was fixed. - Bug in IRT models/Training/Dichotomous models text of theta in plotly was fixed. - Bug in IRT models/Training in plotly a ModeBar was fixed. Major updates - Data upload now allows also ordinal data, min/max values can be specified. - Validity/Correlation structure now allows selection between Pearson/Spearman/Polychoric correlation, Polychoric is default for binary data - DDplot now allows also ordinal items and offers different types of discrimination indices to be displayed - Download of tables (as .csv files) is now provided for standard scores, correlation matrix, traditional item analysis, and for parameter estimates in IRT models. Minor updates - Data upload now better points to "Upload data" button. - Precision was re-set for IRT models/Training tab. Now = 0.1. Changes in version 1.2.8-2 Major updates - gDiscrim() function has been extended to calculate Upper-Lower Index (ULI) also for ordinal items Changes in version 1.2.8-1 Bug fixes - Bug in IRT dichotomous models comparison is fixed. (Reported by Jean-Luc Kop) - Bug in IRT dichotomous models (3PL and 4PL) was fixed. Tables of parameters are rendered even in case that EM cycles terminated without convergence. In such a case SD of parameters are set to NA. - Bug in plotDIFLogistic() function was fixed. Function now returns also item.name argument. - Bug in plotDIFirt() function was fixed. Function now returns also item.name argument. Major updates - Global setting for figure downloads was created. This can be changed in Setting section. Changes in version 1.2.8 (2018-09-28) It includes versions 1.2.7-1 - 1.2.7-7 Changes in version 1.2.7-7 Major updates - Section Reliability was added. - Split-half methods including also Revelle's beta were added - Cronbach's alpha method with CI was moved to section Reliability. - Spearman-Brown formula was added. - New function ggWrightMap replaced wrightMap from WrightMap package. - Section IRT models - Selected R code for Rasch and 1PL models was updated. Minor updates - Reference list was updated. Changes in version 1.2.7-6 Major updates - Section DIF/Fairness - Subsection Logistic IRT Z is deprecated. - Subsection Nonlinear IRT Z is now called Generalized logistic. - Subsection Generalized logistic includes now wide range of models and different specification of DIF type. - Subsection SIBTEST was added, including SIBTEST method and Crossing-SIBTEST method. Minor updates - Reference list was updated. Changes in version 1.2.7-5 Minor updates - Some wording was updated. - Selected R code was updated. - Function theme_shiny is deprecated. New function theme_app() was created and applied for all graphics, including R functions in package. Changes in version 1.2.7-4 Minor updates - Paddings were adjusted. - Acknowledgments were updated. Changes in version 1.2.7-3 Bug fixes - Crashing of the app when upload no data was fixed. Major updates - Data page - Design of data page was changed. - User can now specify encoding of missing values. - Basic summary was added. Changes in version 1.2.7-2 Major updates - Phone version - collapsible menu closes when clicked Changes in version 1.2.7-1 Bug fixes - Plotly files in IRT training section are now displayed with mode bar due a bug in plotly package. Minor updates - Some typos were fixed. - Link to google form for feedback was added into About section. - New R icon was added into footer. Changes in version 1.2.7 (2018-05-04) It includes versions 1.2.6-1 - 1.2.6-12 Changes in version 1.2.6-12 Minor updates - Mirror at Charles University is no longer in use. Changes in version 1.2.6-11 Major updates - Function DDplot() - Limits for y-axis were adjusted when negative values of discrimination are detected. - Warning for negative discrimination values were added. - Reports - Some unused parts of reports were excluded. Changes in version 1.2.6-10 Major updates - UI - Some parts were split from main ui.R file - IRT - Exercises in IRT training were extended. - Data - Bug in data upload was fixed. Now length of criterion variable is correctly checked. Changes in version 1.2.6-9 Major updates - Packages - shinyBS is now required. - Data - HCI dataset was added into package. - Description of dataMedical was improved. - Description of training datasets was improved. - About - Page was improved. - IRT - Helps for interactive exercises were added. Minor updates - Typos were fixed. Changes in version 1.2.6-8 Major updates - IRT - Expected item scores were removed from IRT training NRM. - Interactive exercises were added into IRT training - dichotomous models. Changes in version 1.2.6-7 Major updates - Data - Graded data od Medical data set was added. - In all dataMedical data sets the criterion variable was renamed to StudySuccess. - IRT - Expected item scores were added into polytomous models in IRT training. - Selected R code was updated in IRT training. Minor updates - IRT section - Colors of sliders and curves in IRT training section were adjusted. Changes in version 1.2.6-6 Major updates - Regression - NLR 4P model was added into model comparison. Minor updates - README file was updated. Typos were fixed. - Design of figures was unified. - Download button options were unified. - Selected R code syntax was improved in some sections (IRT only partionally, DIF not). - Selected R code was improved for first 4 sections. - Some wording was improved. Changes in version 1.2.6-5 Major updates - IRT section - Wording for IRT training was added and improved. - Selected R code syntax was improved in some sections. This can be done with txt <- gsub("\n", "", txt) txt <- gsub(" ", " ", txt) txt where txt is R code in "". - theme_shiny was created as a default theme for all figures. Now it is used only partionally. Minor updates - README file was updated. Typos were fixed. Links were added. Hexbin was added. - References were updated. Changes in version 1.2.6-4 Major updates - IRT section - Selected R code was added into polytomous models. - Equations were added into polytomous models. - Download buttons were added into polytomous models. - 4PL model was added. - 4PL model was added into model comparison. - Sliders in IRT training are now colored with respect to line colors. - Regression section - Section is newly organized. - NLR 4P model was added. - Reports - 4PL IRT model was added as an option. - IRT equation is generated based on selected model only. - Unnecessary files were deleted. Some cleaning of files was done. Minor updates - IRT section - Selected R code in model comparison was updated. Typos were fixed. - Regression section - Interpretation of NLR 3P model was extended. - Syntax for NLR 3P model was improved. Changes in version 1.2.6-3 Major updates - IRT section - IRT section is newly organized. - IRT training section was expanded by polytomous models. GRM, GPCM and NRM were added. - Syntax for IRT training was improved. Minor updates - IRT section - Sliders steps are now 0.01 in dichotomous models. Changes in version 1.2.6-2 Minor updates - Numeric inputs were removed from IRT training section. - Selected R code was extended in IRT training section by item information function plot. - References were updated. - Mirror web link was updated. Changes in version 1.2.6-1 Bug fixes - Reports with DDF but none DDF item are now correctly generated. Minor updates - Hexbin - was added into footer. - is now as shortcut icon. - Paddings - in large devices was adjusted. - in small devices was adjusted. - Unnecessary files were deleted. - Reports - Some wording in reports was improved. - PDF report text is fully justified. - Subsections characteristic curves is not visible when none DIF or DDF item is present. - Some typos were fixed. - Some parameters were renamed. The unnecessary ones were removed. Changes in version 1.2.6 (2018-03-02) Major updates - UI for DIF/Fairness section was updated. - References were added into About page. - New design of reports is now available. Design of PDF reports was updated. - Wording in report was improved. - DIF/Fairness section in report was improved. - New hexbin for ShinyItemAnalysis was created and added into reports. - Sample code in DIF/Fairness IRT Lord and IRT Raju was extended to 1PL, 2PL and 3PL models. - Bugs fixing - Wording in regression plots was fixed. - Examples in DIF/Fairness IRT Lord and IRT Raju were fixed. - Function plotDIFirt() now can be also used for only one item. Minor updates - Reference list was updated. - CITATION file was modified. - In Summary page the description for summary table was added. The sample code was updated. - Wording in summary statistics in Summary section was improved. - Description of Session info section in reports was added. Changes in version 1.2.5 Major updates - Correlation plots now offer reordering by hierarchical clustering. Several linkage methods are offered. This is also included in reports. - Confirmation of data upload was added. Also warnings and errors are displayed in case that something went wrong when uploading. - In IRT training - parameter theta (latent ability) is now included - the interpretation for characteristic curves is provided - horizontal and vertical lines for specified theta were added into plot for characteristic curves - selected R code was added. - In multinomial regression bug when only two choices in data are present was fixed. Minor updates - Link for Adela's website was added. - Columns for BIC and its adjusted version were swapped in table for IRT models comparison. Text above table was also updated. - In IRT section dpi of downloaded plots was set to 300. - In IRT training downloaded plots also include legend with parameters. - In all IRT models table with parameters includes all parameters a, b, c, d. - In IRT training equations were moved above interpretation. Changes in version 1.2.4 Minor updates - Reports were mildly updated. - Resolution of figures in reports was increased (now 300ppi) - Dependencies were updated. - Some typos fixed. - Legends of plots were unified and placed in better way for - NLR IRT Z - DDF - function plotDIFLogistic() - Bug in item names for DDF subsection of Fairness/DIF section was fixed. Changes in version 1.2.3 (2017-08-22) Major updates - Report's customized options were extended. Reports now include also dynamic description. - Author and dataset name can be now filled and add into reports. - Dataset difMedical was renamed to MSAT-B according to difNLR package. Its description was mildly updated. - Purification options were added into Nonlinear IRT Z and DDF subsections of DIF/Fairness sections. - Purification options was added DDF analysis in Reports section when customized setting is selected. Minor updates - Size of plots and font in plots was changed for reports. - Plots in HTML report are now centered. - Wording of reports' default option was updated. - References were updated in Reference section and also in README file. - Wording of Spearman test in Validity section was mildly updated. Changes in version 1.2.2 Major updates - Design of reports was edited. - Several bugs were fixed: - Bug when creating report for uploaded data was fixed. - Bug when creating report for data without predictive criterion or DIF group was fixed. - ShinyItemAnalysis now uses plotly package! Minor updates - Design and options of Report page were updated. - Layout of IRT curves training page was mildly changed. There are now two curves in each plot. Also text area input was added and synchronized with sliders. Colour of sliders now correspond to colour of curve. Changes in version 1.2.1 Major updates - Several bugs were fixed: - Interpretation of regression models was fixed. - Interpretation of text about correlation in Predictive validity was fixed. - ShinyItemAnalysis now used also data.table package. - Data page was divided into two subpages to avoid long rendering of tables. when changing datasets. - Function gDiscrim() to calculate generalized discrimination was implemented. - Function DDplot() was edited to be able to handle generalized discrimination. Double slider was added into Traditional analysis page. - Traditional analysis table was extended by adding column with generalized discrimination. This is done also in reports. - Some pages has now their own source code files: - Summary - Summary.R - Validity- Validity.R - Traditional analysis - TraditionalAnalysis.R - Regression - Regression.R - IRT models - IRT.R - DIF/Fairness - DIF.IRT Minor updates - Last three rows in regression comparison table are now bold. - Reference list was extended. - The subpage Distractors of Item analysis page was mildly edited. - Generalized DD plot was removed from Distractor subpage. It is now available on Traditional item analysis subpage. - Item patterns frequency plot is now made with ggplot2 package. - Names of downloaded plots were customized and their resolution was increased and size of text was descreased in some pages: - Summary - Validity - Item analysis - Regression Changes in version 1.2.0 (2017-06-16) Major updates - ShinyItemAnalysis can now handle missing values. Newly there is an option to handle NA values as zeros in scored data. - Bug in WrightMap was fixed. - We have new contributor - Lubomir Stepanek! - New online mirrors were added. - Busy indicator and tooltips were added. - Structure of application was mildy updated. - Datasets dataMedical, dataMedicalkey, dataMedicaltest were updated. Group variable and criterion variable were added. - The newest versions of some packages are now required (difNLR v 1.0.2, difR v 4.7, ggplot2 v 2.2.1, mirt v 1.24, shiny 1.0.3 and shinyjs v 0.9). - DIF/Fairness - Option of item purification was added into some DIF detection methods - Delta plot method, Mantel-Haneszel, Logistic regression with total score, Lord's test and Raju's test. - Default option of p-values adjustments in DIF detection methods was set to none. - Bugs in DIF Logistic IRT Z and DIF Nonlinear IRT Z tables of parameters were fixed. - IRT models - IRT models are now fitted only by mirt package. - Training plots of item characteristic and information curves were added into IRT models section. - Pearson's correlation coefficient between factor scores and standardized total score was added for Bock's nominal IRT model. - Reports - New options for reports were added. By default no IRT models and no graphs of correlation structure are fitted. - Reports are now described in detail. - The settings of reports are now also offered in Report section. - Generation of reports is now done part by part. Once it is finished, the download button is offered. - Message boxs with warning for generation and download of reports were added. - Validity analysis: - Validity analysis of predictive outcome is now included on Validity page in Predictive validity section. - Now it is possible to upload predictive outcome. - Dataset GMAT and GMATtest from difNLR package include now criterion variable which can be used in Predictive validity section. - Functions DistractorAnalysis() and plotDistractorAnalysis() were extended by matching argument. - Item analysis - Barplot of answer selection proportion was added into Distractor subsection. - Diagram of custom discrimination - new version of DD plot - was added into Distractor subsection. Minor updates - CITATION file was added. - Maximum upload size was set to 30MB. - Item names were fixed and unified. - Some typos were fixed. - Readme file was updated. - Reports design was slightly updated. - Code was slightly updated to correspond to new versions of packages difNLR, mirt and ggplot2. - Selected code was updated. - Package dependencies were updated (packages knitr and xtable are now used in report generation). - Minor bug in DIF IRT (Lord and Raju) plots and downloads was fixed. - Designs of footer and navigation bar were updated. - The version for small media was designed. - Weblink for version 1.1.0 was created. - Reference list was expanded and links to online versions of articles were added. Changes in version 1.1.0 (2017-02-03) Major updates - New overall appearance using shinythemes package. - New datasets dataMedical, dataMedicalkey, dataMedicaltest were added into the package. - New dataset is offered in example datasets inside the shiny app. - Bug in Distractor plots for uploaded data was fixed. - Calculation of Success Rate in Standard Scores was fixed. - Differential Distractor Functioning detection method was added. - Syntax of difNLR functions was edited to response to new version of package. - Function plotDIFirt() was updated. - Correlation structure section was added including correlation plot and scree plot. - Pages structure was changed and updated. - IRT models are now fitted by mirt package. - Subsection IRT model selection was added into IRT models section. - Wrigh maps were added for Rasch and 1PL IRT model. - Bug in data upload was fixed. - Subsection regression model selection was added into Regression section. Minor updates - Typos were fixed. - Selected R code was updated. - Description of data including uploading own data set was extended. - Description of Standard scores was added. - Lower panel was updated. - Rasch model was added into IRT models section. - Pearson correlation coefficient of factor scores and Z-scores were added for IRT models - Reference list was expanded. - Bug in multinomial regression for scored data was fixed. - Bug in DDF for datasets withou DIF items was fixed. - Selected code was updated. Changes in version 1.0.0 (2016-11-25) Major updates - New graphical representation via ggplot2 package - Cronbach's alpha was added - Distractor analysis was extended - More DIF detection methods were added Mantel-Haenszel test, Logistic regression, Lord's and Raju's statistic for IRT models all are now used via difR package - Delta plots were extended: fixed and normal threshold are now allowed - Multiple comparison correction methods were added into DIF detection procedures - More examples of datasets were added from difNLR package - References were added - New functions were added into the package: - DDplot() function for graphical representation of difficulties and discriminations. - DistractorAnalysis() function and its graphical representation plotDistractorAnalysis(). - plotDIFirt() and plotDIFLogistic() functions for plotting characteristic curves based on IRT models and based on logistic regression.