Latent variable analysis has become more and more popular among social scientists since the extension of factor analysis for metric variables to categorical variables. In the literature different approaches have been developed for handling categorical data and they have not always been successful either because model assumptions do not hold or because they are computationally heavy.
The talk will give an overview of available methodology for handling categorical variables within latent variable analysis. Researchers have a lot of flexibility since there are many models to choose but there is little guidance available on which model to choose.
The critical issues related to categorical data analysis are in my view the amount of information to be used for the modeling, the effect that that has on the goodness-of-fit of the model and computational issues that are important for the practitioner. The goodness-of-fit of the models remains mostly an unsolved problem still under investigation.
The talk will give emphasis to the methodological advantages of the different methods available, goodness-of-fit issues as well as the latest software developments.