The Multidimensional Item-Component Rasch-Model (MultiRa) is a probabilistic test model. It assumes that responses of a test or questionnaire can be explained by several latent traits. In this sense it is a probabilistic alternative to factor analysis.
More Information on this Model can be found in the Manual, which is included in the program download.
MULTIRA v1.65 (by C. H. Carstensen and J. Rost) is a Computerprogramm that implements an algorithm for the multidimensional item component Rasch model and other Rasch models. Actually, the progam includes:
- Person and Itemparameter estimates according to the MultiRa-Model
- Joint maximum Likelihood and conditional Maximum Likelihood Estimation
- Modelfit measures including
- Martin Loef Test on Item Homogeneity
- Residual Mean Square Item Fit
- Bootstrap GoF
- MKAT Algorithm by J. Rost for Raschs multidimensional Model (1961)
The MULTIRA-Program is free available as beta-release. To obtain further informations and updates, please tell us your E-Mail adress.
- 20.06.03 MULTIRA v1.65: minor bugs and causes for program crashs removed.
- 27.11.01 MULTIRA v1.63: A bug in the accelerator was removed.
- 05.09.01 MULTIRA v1.62: CML estimation is "accelerated" by the Aitken Accelerator. The missing data option has been checked further and is considered as working well. Therefore only v1.62 is available for download.
- 01.07.01 MULTIRA v1.61: The computation of WLE (Warm estimates) has been revised; it is faster than previously and numerically more stable. The Standard errors for WLEs are computed correctly now.
- 11.06.01 MULTIRA v1.60 handles missing data estimating parameters using the JML algorithm. Also an option was implemented to fix item parameters. Due to the complexity of the change to handling missing data and a yet only short testing phase two versions of MULTIRA will be downloadable, the previous one v1.51 and the new one v1.60.
- 24.4.01 MULTIRA v1.51: The number of scores for within item multidimensional models using the CML algorithm is now computed correctly (formerly an upper limit was printed).
- 13.12.00 MULTIRA v1.50: The algorithm MKAT (J. Rost) performing CML estimation of the multicategorical Model of Rasch (1961) has been revised and is available now. The algorithm for the dichotomous LLTM has been revised and is available, as well.
ML estimators for the person parameters can be calculated for person with extreme scores by reassigning them "very" small rsp. large non-extreme-scores.
The MULTIRA Manual is available as Adobe PDF file, now. It is still in german language only.
- 13.7.2000 The Martin-Loef Test for Item homogeneity can be performed for ordinal data and more than two item groups (see Carstensen, Claus H., 2000. Mehrdimensionale Testmodelle und deren Anwendung in der pädagogisch-psychologischen Diagnostik. Kiel:IPN).
In MULTIRA v1.44, some bugs were removed: the re-categorization option is working properly, now.
- 01.03.2000 In program versionv1.43 some details have been added to the CML estimation procedure. It is now fully functional for multidimensional tests.
The number of subjects excluded because of extrem scores is now correct in multidimensional tests
- October 99 The program version v1.42 comes without some errors in the calculation of the Chi-Square statistics on pattern frequencies. With CML-estimates, person parameters are no longer used to calculate those statistics
- 21.4.99 The new program version v1.40 offers conditional Maximum Likelihood-Estimation (CML) for the MultiRa-Model.
- September 98 A manual in german language is available.
- April 98 In the program version v1.23 a norming constraint for facetted design is implemented.
List of known bugs
- Using blanks in filenames can still cause troubles with filenames.
- The algorithm MKAT is limited in the number of categories and items that can be analyzed.
The english program (version 1.65) is packed into the Archiv >MULTIRA_setup_eng.exe(2.4 MB)<. To install the program run "MULTIRA_setup_eng.exe".
The (german) Online-Manual(424.6 KB) (>mmanual_de.pdf<) offers further informations.
Link zur deutschen Homepage mit Download.