GTT can read in data from plain text-files in free or fixed-width formats (Project⇒New). Positive and negative responses are assumed to be numerically coded as 1 and 0. The default code for missing values is 9. Only one line of data per individual is allowed. In case the data file contains header lines or columns containing individual IDs, it is possible to skip lines at the beginning and a of columns in each row. After the data have been imported, they are shown in the data window. To each item there belongs a check box for item inclusion and reverse item coding. The data can be stored as a GTT project file (Project⇒Save as) that allows immediate later access to the data (Project⇒Open). 

The CTT menu can be used to perform item analysis based on classical test theory. Specifically item difficulty and item discrimination can be calculated via (CTT⇒Item analysis). In addition the individual sum scores are available along with a frequency table of the observed scores as well as a table relating sum scores to the T-, Z- and C-norms. These norm values have means and standard deviations according to the following table:

The IRT menu can be used to perform item analysis based on item response theory. Before starting item calibration the user can select among the 1-, 2-, and 3-PL models. The statistical approach to item calibration is mainly marginal maximum likelihood (MML) using the EM Algorithm. For the 1-PL model the conditional maximum likelihood approach (CML) is also available. The calibration is run by clicking IRT⇒Run. To fix the scale for the item parameters, MML estimation assumes standard normally distributed abilities, and CML estimation scales the items difficulties such that their mean equals zero.

The result of a model fit is shown in a new window. It is possible to print the output and to save it in different formats:

Calibration settings (IRT⇒Settings) can be modified by the user. This is useful if item calibration results in warnings or error messages. Specifically, users can specify:

• Max EM iterations: Maximum number of EM iterations.
• EM termination criterion: Maximum absolute change of any single item parameter between successive iterations.
• Max Newton iterations: Maximum number of Newton iterations when performing the M-step.
• Newton termination criterion: Maximum absolute item parameter change between successive iterations of the M-step.
• Starting values: Estimates will be calculated by fitting logistic regression models for each item (option Estimate). It is possible to specify starting values using the Custom option.
• Number of nodes: Number of rectangles for numerical integration in the E-step.
• From: Lower integration interval limit of the E-step.
• To: Upper integration interval limit of the E-step.

GTT can display the item characteristic curve (ICC), test characteristic curve (TCC), item information and test information. Plots⇒Run will display the plots for which the corresponding menu entry is checked. Each plot provides controls for selecting which graphs to show, which horizontal and vertical ranges to show, grid intervals and whether to show labels.

After fitting a model to the data, ability estimates can be calculated (Abilities⇒Estimate). Users can select among the maximum likelihood (ML), maximum a posteriori (MAP) and the weighted likelihood estimate (WLE). In addition, the following settings influencing the numerical computations can be changed:

• Max iterations: Maximum number of iterations.
• Termination criterion: Maximum absolute item parameter change between successive iterations.
• Automatic starting values: GTT can automatically select a starting value based on a rough guess of the ability estimate.
• Starting values: Users can supply a starting value, which will be used as a starting value for all individuals.
• Step size: Factor by which the change in the ability estimate between successive iterations can be shrunk. (A factor less than 1.0 maybe useful in cases of non-convergence.)

GTT computes Yen's Q₁ statistic to evaluate item fit. For the 1-PL model the infit and outfit statistics are also available. Since item-fit statistics depend on ability GTT allows users to select the ability estimate on which the statistics are computed.

For Q₁, two modes of ability-based grouping of individuals are available:

• Equal group size: Users can select the group size.
• Equal probabilities: This setting assumes normal distribution of abilities and uses the distribution's quantile function to determine ability intervals with equal expected frequency of occurrence. Small groups (Min. group size) will be removed and values will be distributed among adjacent groups.

GTT has been developed at the University of Giessen. Program authors are Christof Schuster and Laurens Berthold.  GTT has been developed using Qt, Eigen and Boost. Qt is licensed under the LGPL 2.1. Eigen is licensed under the LGPL 3.0.