2-5 September 2025

The course “Latent class, Markov and hidden Markov models with applications in R” aims at introducing the main basic theory of the latent class models for cross-sectional data and Markov and hidden (or latent) Markov models, that are tailored for the analysis of longitudinal data. Basic and advanced methods to develop these models are explained also when explanatory variables are collected. An insight of the underlying assumptions, which are based on stochastic processes, is provided as well as guidelines on how to apply these models. Response variables can be quantitative or categorical, and even incomplete for some subjects. The inferential procedures are explained in detail.  Applications from different fields and exercises are developed by using the R environment. The course also contains interactive practice sessions during which participants can bring up their own concrete use cases and question sets.
The course is intended for students from academia and for practitioners from industry and consultancy with little or no background in probability who wish to apply latent class, Markov and hidden Markov models. Basic familiarity with R is desirable but not mandatory.

The course will be based on some chapters of the following book Latent Markov models for Longitudinal Data. Empirical illustrations and exercises will be carried out using the R libraries MultiLCIRT and LMest.

After completing this course, participants will:

1.    Be familiar with the basics of discrete latent variable models.
2.    Be able to estimate and interpret the results of various latent class models with covariates.
3.    Be able to estimate different Markov and hidden Markov models with continuous and categorical longitudinal data, including cases with missing values.
4.    Know how to perform model selection and estimate the uncertainty of model parameters using bootstrap procedures.
 

Tuesday 2 September - 2-5 pm Berlin time

Course overview, practicalities. Introduction to latent variable models with a focus on discrete latent variable models. Basic features of the finite mixture and latent class model with reference to the estimation method and to the Expectation-Maximization (EM) algorithm. Applications with the R library MultiLCIRT using RMarkdown.


Wednesday 3 September - 2-5 pm Berlin time

Introduction to longitudinal data. Basic features of the Markov model for continuous and categorical longitudinal data with and without covariates, estimation and model selection procedures. Applications and exercises with the R library LMest using RMarkdown.


Thursday 4 September - 2-5 pm Berlin time

Hidden Markov model formulations and extensions with multivariate data, covariates and more complex data structures. Recent case studies and applications. Exercises with the R library LMest.

Friday 5 September- 2-5 pm Berlin time

1. Good practices: how to report results, corrections of the assigned exercises, …
2. [topics of interest to the audience / general discussion]