After completion of this course, students hould be able to
- explain the concept of local independence in the context of latent variable modeling (LVM) and its importance from a substantive and statistical point of view.
- execute a basic latent class model including model selection, model interpretation, and classification.
- to reproduce output of LCA by explaining or solving the necessary computations by hand.
- describe the most relevant extensions of the simple latent class model, including LC-models allowing for local independences, models with multiple latent variables, ordered latent class models, and models with covariates.
- solve classifications and contrast different approaches for using latent class classifications in subsequent analysis (including the three-steps approach).
- describe similarites and differences between latent variable models for discrete and continuous latent variables.
- describe popular item response theory (IRT) models as latent variable models that assume a continuous latent variable.
- demonstrate similarities and differences between latent variable approaches using the latent variable measurement model and network psychometrics.
- state the role of latent variable measurement models for substantive research and questionnaire construction.
- interpret the basic results of LVM analysis (parameters, GoF, test information) of dichotomous or polyomotous questionnaire data and examine the adequacy of the instrument for practical purposes envisaged.
- compute the (conditional) response probabilities under different LVMs, including models for dichotomously and polytomously scored items..
- apply latent variable models that includes observed covariates to test substantive hypotheses about the attributes under consideration and to evaluate the quality of tests and questionnaires (e.g., measurement invariance and detection of item bias).
- describe basic network theory and methodology to construct and analyze psychological networks.
- apply networks by encoding networks using adjacency matrices and fitting the Ising model.
- interpret network structures in terms of edges, nodes, centrality and other characteristics
This course provides an introduction to latent variable measurement and psychometric networks in the psychological and social sciences. The models to be discussed include latent class models, linear factor models, and latent trait models (item-response theory models), and the Ising netwerk model. These models are routinely applied in various fields of the social sciences. We will focus on the conceptual foundations of the models, discuss the basic model and its generalizations or special cases, and practice different applications to real data sets. The relevant software in this course includes SPSS, R, LatentGOLD, and IRTPRO (student version).
Type of instructions
14 Lectures (2 hours) and 7 lab sessions (2 hours)
Attendance at the lab sessions is strongly advised. Students will not be able to complete the individual assignments without the training in the practicals of the course.
Type of exams
This course will have an open-questions exam. A student passes the course if the final grade is 5.5. or higher.