SSI, Inc. offers a technical support service via e-mail. This service is only available to registered end users of the SSI software products. Click here to view SSI’s policy and procedures for obtaining technical support.
This service applies to problems in using the software and does not include statistical consultation.
Documentation
FAQS
Can IRTPRO spreadsheet files (*ssig) be exported to SPSS, SAS, STATA, etc. formats?
This feature is available in IRTPRO™ 4 and is accomplished as follows:
Use the File, Open option to open an .ssig file
Use the File, Export option to export this file to a different format
...
“-Run” is used to open IRTPRO™ and run the analyses described in a single .irtpro command file. In the Command Prompt window, enter a line of the form:
...
In commands typed in the Command Prompt window, or in .bat files executed there, double quotes are required around path/file names that contain spaces.
Several lines of the form(s) above can be created one after another in a file with the extension .bat; then that file can be double-clicked, or invoked in the Command Prompt window, and the lines will be executed one after another.
The .irtpro files can either be created by the IRTPRO™ GUI, or appropriately edited versions of such files, or (possibly) even created by some user-written special-purpose software.
It may be useful to select the -irt.txt, or other .txt file, output in the commands in the .irtpro file, to yield ASCII files that can be post-processed after a sequence of “batched” runs.
-RunList
“-RunList” RunList” is used to open IRTPRO™ and run the analyses described in a list of .irtpro command files; the paths/names for the .irtpro command files are specified as lines of a .txt file.
...
Like Multilog, IRTPRO™ uses maximum marginal likelihood estimation, which involves no priors, when doing item parameter calibration, unless the user instructs otherwise. The user may select item parameter priors through the Priors tab in Advanced Options dialog box. Or, alternatively, write in the Priors syntax section. Three kinds of prior densities are supported: Beta, Log-normal, and normal. These univariate priors can be imposed on any item parameters. However, for the guessing parameter, the normal prior actually means logit-normal on the logit of guessing probability. The log-normal prior is mostly meant for the slope parameter in unidimensional models. The beta prior is supported since people are familiar with the Beta priors in Bilog BilogMG and Parscale. Due to the general multidimensional and hierarchical model on top of which IRTPRO™ is built, we chose not to assume default priors as does Bilog BilogMG or Parscale (which have more limited focuses to unidimensional analysis and can therefore make more assumptions about the user’s intent). Hence, by default, no priors are imposed on any parameter in IRTPRO™. This can cause problems if the 3PL model is used. Unless sample size is huge, a logit-normal or beta prior on the lower asymptote parameter should be imposed at the very least.
...
When using IRTPRO™ to score individuals in a dataset using a set of item parameters from an external file, one of the items might not have any individuals in the dataset that responded to an item in, for example, the first category. IRTPRO™ will not score correctly in this case. To solve this, one could add a fake person to the dataset with a response of 1 for that variable so that it would recode people correctly , and then drop the fake person after scoring the dataset. This is an annoyance, but there is at present (and for the foreseeable future) no easier solution. When this happens to us, or when we fear it might happen, we just add as many “fake persons” to the dataset as there are response categories. For five-category items that’s five: If the codes are 1-5, then there’s one that’s all 1s, the next is all 2s, etc. up to all 5s.
...
The –prm.txt output file is a tab-delimited ASCII .txt file that contains one line for each item and/or group. It is an output file from item calibration runs , and may be used as a parameter input file for stand-along scoring runs.
The tab-delimited format makes it easy for users to edit, or store, by opening it with, say, Excel.
The one-line-per-item format makes it easy to “mix and match” lines from multiple runs of the program to “assemble” new tests—new combinations of items one may want to score.
Each line contains a variable number of fields, depending on the number of dimensions and the item response model:
...
An example, that illustrates this problem is as follows: when the Analysis, Models tab is displayed, it does not show the Constraints…or DIF… buttons.
To fix this, experiment with your display settings and specifically the scale (which should be made smaller). Proceed as follows: Right click on your computer monitor screen (not on an icon) and select the Display settings option. Select the smallest scale value available. If this does not solve the problem, try various sizes for your display resolution
...
Definitely acceptable characters are: ‘0’-‘9’, ‘a’-‘z’, ‘A’-‘Z’.
Special characters that should work (not fully tested: ‘@’, ‘_’, ‘$’, ‘#’
Definitely do not use: ‘-‘ (dash or hyphen) , ‘:’(colon), ‘;’ (semi-colon). ‘-‘ and ‘:’ have special meanings in the syntax: they are used as range specification. Only ‘-‘ is used in current version but both will be used in version 6.
Questionable: Not recommended (not tested, at own risk): all forms of brackets ({, }, [, ], (, ), <, >), path separators (\, /), miscellaneous symbols such as grave accent (`), tilde (~), circumflex (^), percent (%), ampersand (&), asterisk (*), vertical bar (|), question mark (?), space (enclosed by a pair of matching quotes).
Single and double quotes are used to quote string of characters that include special characters including space, therefore quotes characters will never be treated as part of the names.
International non-English characters are not supported.
...
The difference between R or r (and B or b) depends on the absolute magnitude of the local dependence. The printed numbers are standardized Pearson X2s to ensure that all cells are comparable regardless of the size of the underlying two-way contingency tables (a 2×2 table vs. a 5×5 table would have different expected degrees-of-freedom for the X2). If the standardized X2 exceeds 3.0 in absolute terms, which is substantial, capitalized R or B would be printed. Otherwise, lower case letters are used.
...
In this example we illustrate how to set up fixed item calibration with IRTPRO™, wherein some of the items have been previously calibrated and come with item parameter estimates that are assumed to be known. New items with unknown and freely estimated item parameters are also present in the analysis. This example also contains a mixture of IRT models (3PL and Graded), which is common in mixed format educational tests. (download complete example)
What is the difference in parameterizations used for common IRT models between legacy IRT programs and IRTPRO™?
...
If providing guessing values for the 3PL, the values provided to IRTPRO™ will need to be logit-guessing, not typical g-parameter values. This PDF provides additional details on the parameterization of the 3PL (and GPC) in IRTPRO™.
...