Central assumptions in psychological, social, and economic evaluations are that all items in a questionnaire have the same discrimination power, unidimensionality of the latent trait, and measurement invariance of the scale. We briefly revise latent Markov models for modeling data arising from repeatedly measured items. We then focus on the concept of measurement non-invariance. We define different notions of differential item functioning in the context of panel data. A simple model selection approach based on the Bayesian information criterion (BIC) is argued to successfully identify the correct measurement structure. We show the practical relevance by means of an extensive simulation study, and illustrate its use on two real-data examples from the social sciences.
This is joint work with Antonio Di Mari, Francesco Dotto, Antonio Punzo
about the speaker
Alessio Farcomeni is Professor of Statistics at University of Rome “Tor Vergata”. His main research interests involve multiple testing, quantile regression, survival and longitudinal data analysis, manifest and latent categorical data, population size estimation, robust statistical analysis. He is currently author or co-author of more than 300 papers in international peer-reviewed journals, two books, several software packages. He has extensive experience in collaborating with researchers from several fields, including economics, medicine, biology, ecology, engineering. He is listed as one of the top Italian scientists by VIA-Academy and by Ioannidis et al. (2020, 2021).