Maria Iannario - Latent trait models for ordinal categorical data

Latent variable models are used to explain the interrelationships between set of observed variables by a smaller set of latent variables using survey analyses.
Different methods have been developed for the analysis of both single and multiple observed items. In the former case, assuming that the discrete response score is ob- tained by grouping the latent variable surrounding the individual choice into classes of values by means of cut points, and assuming that there is a relationship between the cumulative function of the ratings and the subjects variables. For the second case, a generalisation regarding the Item Response Theory framework is taken into account.
In recent years, the departure from these usual practices has been discussed in order to take into account the heterogeneity or variability in the process of response and possible response styles.
The seminar will cover these rich classes of ordinal response models, including recent developments and the possible study of partially ordinal scales where the ’don’t know’ option is provided.
An analysis of case studies is carried out to evaluate the performance of the methods and to demonstrate the valuable support they provide to the process under investigation.