Latent class analysis is the name social scientists originally gave to the study of “mixtures of Bernoulli models” - the search for hidden groups in categorical data. Since then, the term has grown to mean almost any kind of model in which there are thought to be different groups, and the problem is that we do not know which groups. Examples are “latent profile analysis”, “Gaussian Mixture Modeling”, “mixture structural equation modeling”, “model-based clustering”, Hidden Markov modeling, and many many more. Latent class(-type) models have found application in market segmentation, ideal point modeling in political science, diagnostic test evaluation without a gold standard, probabilistic record linkage, disease stratification, image recognition, and student mastery models (CDM), to name just a few.

 

In this course, I will assume you know something about statistics, including linear and logistic regression, as well as programming in R. Starting from that assumption, we will work to understand the goals and promises of latent class modeling, broadly understood, as well as its inner workings in simple cases. You will learn to formulate and work with several variations of latent class model, as well as:

• What such models could be used for;

• How to evaluate and compare them;

• How they can be interpreted and what you cannot get out of them;

• What software options there are.

We will use R throughout because it is the most accessible option. However, using R for latent class modeling can be challenging: It requires navigating different, sometimes changing, R packages of varying quality. Moreover, for some techniques in use within (among others) the social sciences, no ready-made R package is available at all, and the user is required to get “creative”. For this reason, you will probably enjoy this course more if you feel rather comfortable with R and are not easily frustrated by error messages. (Chocolate may be provided to lower stress levels during key parts of the course).

 

Organizational structure of the course

Morning: Lectures 3 hours.

Afternoon: Tutorial 3 hours.

Software and hardware requirements

Participants will need to bring their own laptops with R (latest version) and R studio (or any other R IDE; latest version) installed. Participants will need to be able to install packages (i.e., have the according admin rights on their machines) and connect to a WIFI (free WIFI access will be available in the classroom). Participants are expected to be comfortable with R and not afraid of error messages.