The Rasch-online project aims at proposing to students or researchers a way to perform
analyses in the field of Item Response Theory (and more specifically in the Rasch Paradigm) on their
own data, in a user-friendly manner, without complex handling of the data.
The proposed programs constitute an extension of the Anaqol project
(http://www.anaqol.org) that includes SAS
macro-programs or Stata modules. These developments are supported by the team of subjective variables
measurement of the SPHERE unit (http://www.sphere-nantes.fr).
This unit is part of the University of Nantes (France).
The team of this project is composed of:
- Jean-Benoit Hardouin (head for the statistical programming)
- Tanguy Le Néel (IT)
- Myriam Blanchin (statistical programming)
- Jean-François Hamel (statistical programming)
- Véronique Sébille (Director of the SPHERE unit)
Some previous work has shown that using the classical formula developed for manifest variables
(assumed normally distributed) for power and sample size determination when a Rasch model is
intended to be used for analysing PRO data is inadequate.
BMC Med Res Methodol. 2010 Mar 25;10:24. doi: 10.1186/1471-2288-10-24.
It leads to overestimation of power and underestimation of the required sample size.
Hence an adapted methodology, raschpower, has been developed in the context of latent variables
and IRT modelling.
The raschpower module provides an estimation of the power for the comparison of
Patient-Reported Outcomes data using IRT (Rasch model) in cross-sectional and longitudinal
studies on the latent variable measured by an IRT model.
Mokken Scale analysis
Allows analysing a dataset using Non parametric IRT models, in particular Mokken models. This module computes:
- The Loevinger's H coefficient of scalability
- The Loevinger's H coeffient for each item
- The Loevinger's H coefficient for each pair of items
- Indices to check the monotonicity assumption of the items
- Indices to check the double monotonicity assumption of the items
Sign-in for a free account