Change log

All major and minor updates, changes and bug fixes are listed here starting with version 2.0. To take full advantage of the features in jMetrik, be sure you are using the most current version. Starting with version 2.0.1, jMetrik will automatically check for updates on startup. If you find a bug, please send an email to


April 19, 2014




Bug fix

  • Importing delimited files resulted in a lexical error in a few rare cases. The CREATE TABLES statement of the import procedure (not visible to users) was modified to use double quotes around column names. This change eliminated lexical errors due to the character encoding of the names.

March 28, 2014




Bug fix

  • A development version of the Advanced and Basic Item Scoring Dialog was accidentally included in version 3.1.1. The development version was not complete and would result in an error but produce no error message. The developmnet version of these scoring dialogs was removed and a new version number issued.


March 25, 2014




Minor bug fixes

  • A missing response in the first row of data prevented item scoring from being initialized. The result was a null pointer exception during a Rasch Models analysis. Problem corrected by initializing the scoring in a different place.

  • Use of the linear transformation option in a Rasch Models analysis caused a null pointer exceptions because the array of item standard errors had not been initialized before the linear transformation was applied to it. Problem corrected by switching the order of when the standard errors are computed and when the linear transformation is applied.

  • Collapsing categories in the Rasch Models analysis resulted in items being dropped even when there was adequate data. The problem was due to getting the item category count from the number of response options instead of the number of score categories. Problem corrected and items are no longer dropped when collapsing categories.

  • Running IRTPlots with more than 100 items would produce an error because the progress bar did not initialize correctly. Problem corrected and a check was added to all instances of the progress bar to make sure progress values were between 0 and 100.



January 14, 2014


Rasch Models refinements and bug fixes

  • Joint Maximum Likelihood Estimation (JMLE) algorithm improved. Calculations complete in less than half the amount of time as version 3.0.

  • JMLE now allows for person centering.

  • Estimation with fixed common items no longer centers estimates about the item difficulty mean. Values now agree with WINSTEPS software ( IMPORTANT: This change will result in different difficulty estimates for items that are not fixed during estimation.

  • Extreme score adjustment no longer occurs in a separate phase. Extreme score adjustment occurs during main estimation routine. Values now agree with WINSTEPS. IMPORTANT: This change may result in different estimates for extreme items.

  • Standardized outfit (Std. UMS) revised to have bounds on the values for small samples. Values now agree with WINSTEPS. IMPORTANT: This change will result in different values for Std. WMS.

  • (bug fix) Partial credit and rating scale model threshold standard error calculations corrected. Values now agree with WINSTEPS. IMPORTANT: This change will result in different standard errors for threshold parameter estimates.

  • jMetrik uses PROX estimates of item difficulty, threshold, and person ability as starting values. These values have been refined in version 3.1. For almost all items and persons, this change will not affect the results. However, because of the new starting values, you may see small differences in one or two estimates. Any differences will be small. They are due how the convergence criterion is satisfied. The table below shows the maximum change in logits for version 3.0 and 3.1. In all cases, the convergence criterion of 0.005 was satisfied. However, notice that the final maximum change in logits is different between the two verison of jMetrik. This result may lead to small differences in the some final estimates. If you set the convergence criterion to a very small value (e.g. .00001), no differences in estimates between the two versions will be apparant.

    Data 3.0 3.1
    EXAM1 0.004625 0.00495
    EXAM2 0.004697 0.00312
    EXAM3 0.003246 0.00242

New analysis features and refinements

  • Added MLE, MAP, and EAP person scoring methods
  • Added linear interpolation to score tranformation table in test scaling methods. Procedure now provide raw to sale score conversion values for scores that are not observed in the data file.
  • Added "Score as zero" option to DIF: Mantel-Haesnzel procedure. This new option is ideal for tables with planned missing data.
  • (bug fix) Frequency tables no longer sorting values in lexicographic order. Now using natural ordering.
  • Extreme item flag added to databse output in Rasch Models procedure.
  • (bug fix) Kelley score caused an error in Test Scaling procedure. Problem corrected.

New interface components and modifications

  • Enhanced preferences for plots and charts. You can customize line styles and colors. Plot preferences moved to the Preferences dialog.
  • Fixed common item calibration now recognizes polytomous items
  • Added button for closing all tabs
  • Added check for item parameter table in IRT linking dialog
  • Added edit menu items for cut, copy, and paste in text files
  • Dialogs for plots changed to reflect new location of plot preferences
  • Added command menu to facilitate running analyses from a command file


April 12, 2013 3.0.1 Changes to installer and minor bug fixes.
-DIF analysis now recognizes numbers for group codes
-Prevent user from deleting active data set
-Robust z recognizes Rasch model items
April 8, 2013 3.0.0 Major rebuild of jMetrik. It includes new database features, methods of anlaysis, and revisions to the interface. Done
TBA 2.1.1 Refinement: Completely separated psychometrics library from jMetrik library and made pscyhometrics library available online.  
    Refinement: Added Likelihood cross-validation bandwidth computation to kernel class  
    Refinement: Added Least-squares cross-validation bandwidth computation to kernel class  
    Bug fix: Nonparamtric OCC DIF analysis not running from syntax.  
    Bug fix: Textbox sizes in nonparametric occ dialog (not displaying properly on Mac OSX  
    Bug fix: DIF analysis by nonparametric curves not running from syntax due to a misspelled command. Fixed
August 7, 2011 2.1.0 New Feature: Item response theory plots Added
    New Feature: Added option to data import that allows user to specify the number of rows of data that are scanned to determine data type. Added
    Refinement: jMetrik command syntax simplified and now allows for arguments with spaces. Added
    Refinement: New variance estimator for the standardized mean difference statistic in the Mantel-Haenszel DIF analysis. CMH DIF dialog box updated accordingly. Added
    Refinement: Import/export modified to handle quoted text. Added
    Refinement: Import/export dialogs improved Added
    Refinement: Scoring and grouping commands changed to be more readable. Added
    Refinement: Adjusted variance in Rasch scale quality statistics is now constrained to only provide nonnegative variance estimates. This change prevents NaNs from being displayed for adjusted variance, separation, and reliability. Added
    Bug Fix: The option to save person estimates in a Rasch analysis produced an error because the analysis attempted to update the GUI from outside the EDT. This bug was not related to any calculation. It only affected the performance of the interface. Bug corrected by making the call to the GUI update within the EDT. Fixed
    Bug fix: In IRT equating, transformed item parameter variables and person parameter variable would add a "t" to the end of the original variable name. If the original name exceeded 10 characters this "t" would be trimmed off resulting in a nonunique item name. A database error would result. Corrected by adding the "t" to the beginning of the original variable name. Fixed
    Bug fix: An error would occur when users selected the save item parameters option in a Rasch analysis but did not provide a name for the table. Users now prompted for input if name of item output table is omitted. Fixed
    Bug Fix: A "Null Pointer Exception" was thrown when using the "Item Deleted Reliability" option when you have item groups specified. This occurred because the program attempted to do an item deleted reliability analysis for the part-test reliability estimates when no such functionality existed for the part-test reliability estimates. Fixed
    Bug fix: If a user deleted a workspace outside of jMetrik or used a workspace located on a portable hard drive that is no longer connected, Jmetrik would freeze on startup if that workspace was the last workspace. jMetrik will now load the DefaultWorkspace if a workspace cannot be found. If the DefaultWorkspace does not exist, it will be created. Fixed
April 23, 2011 2.0.1 New feature: jMetrik automatically checks for updates on startup Added
    Bug fix: Open and save dialogs would not remember last location. Fixed
    Bug fix: DIF analysis frequency tables not displaying. Fixed
    Bug fix: Rasch item start values setting all start values to last item in database. Fixed
    Bug fix: Running Rasch analysis from syntax with the option of saving item parameter values would save the UCON bias corrected values in the database. This behavior was due to the default options in the Rasch analysis command. Fixed by changing command defaults. Fixed
    Bug fix: Item names not displaying on nonparametric curves graphs. Fixed
    Bug fix: x-axis option for normal scores not working for nonparametric curves Fixed
    Bug fix: Item names not displaying on DIF nonparametric curves graphs. Fixed
    Bug fix: x-axis option for normal scores not working for DIF nonparametric curves Fixed
April 5, 2011 2.0.0 Major revision to jMetrik. New features include test scaling, ranking, item response theory (IRT), and IRT equating. jMetrik now uses a new syntax/command system. Database functioning improved for better performance and reliability. Graphs changed. Memory management improved.