Analysis and Modeling
of Complex Data
in Behavioural and Social Sciences

Anacapri (Capri Island, Italy), 3-4 September 2012
Joint meeting of Japanese and Italian Classification Societies

Satellite meeting of COMPSTAT 2012


We regret to announce that Prof. Reinhold Hatzinger has unexpectedly passed away.

For this reason, the Tutorial on "Extended Rasch Modeling in R" (September 2, 2012) is canceled.

We will contact all the registered participants for the reimbursement of the fee.

Please do not hesitate to contact us for any need.

We are glad to announce that on September 2 it will be held a tutorial on:

Extended Rasch Modeling in R

The instructor is R. Hatzinger (Vienna University of Economics and Business, Austria).

The location is the meeting place, Villa Orlandi, in Anacapri.

Registration is required. Dealine 30 June 2012. Only 30 scholars will be admitted. A minimum number of 25 scholars is required. If by 30 June this number is not achieved the tutorial will be cancelled and fees will be reimbursed.

Selection will be done according to both a first-come, first-served principle and CV evaluation. Please after the registration send an e-mail to with a short CV.

Ph.D. students and young scholars will be preferred.

Registration fees are 70 euro; a reduced fee of 50 euro is applied for students and for meeting attendants.
Registration fees include materials and coffee break (lunch is not included).

A personal laptop is required.

About the tutorial

The Rasch model (RM) is a prominent member of the family of IRT models which specify a functional relation between the probability of a certain response and some parameters describing characteristics of respondents and the situation where the responses take place. Usually, the Rasch model is applied in a testing scenario where the subjects are characterized by their ability and item difficulties represent the situations. The basic Rasch model was formulated as am measurement model for dichotomous items with some important properties such as uni-dimensionality of the latent trait, sufficiency of the raw scores, local independence, and parallel item characteristic curves (ICCs). These properties together with conditional ML estimation allow for
separability of person and item parameters and specific objective comparisons
between subjects or items.

The basic RM may, arguably, be too restrictive for practical testing purposes. Various extensions have been proposed pulling in different directions. Some of these are (i) models for ordinal rather than binary responses, (ii) reparametrizations of situation (item) specific parameters, the so-called linearized RMs, and (iii) introducing additional parameters for discrimination, guessing, and tiring or carelessness, and (iv) explanatory item response models. Whereas (i) and (ii) rely on conditional ML methods, (iii) and (iv) use random effects by specifying marginal distributions (mainly) for subjects.

The tutorial will introduce the underlying theoretical basics of these models, demonstrate properties and applications using some datasets from diverse disciplines, and present some R packages (eRm, ltm, and lme4) that cover analyses using these models. There will be practicals, where the participants learn to apply these methods. A website will be set up, to provide materials for the tutorial, e.g., course slides, some R code, and data sets.

The participants are expected to bring their laptops, with the latest version of R (most likely to be R-2.15.x) and the latest versions of the packages eRm, ltm, and lme4 installed.

For further information, please contact