# What is Latent Structure Analysis

Nick Shryane, ISC

Sometimes the things we want to study can’t be observed directly, in which case we can call them ‘latent’ variables.

The classic example of a latent variable is intelligence, which we infer as the reason why some people do better and some worse at abstract reasoning tests. Other latent variables in wide use in social and psychological science are attitudes, beliefs, etc.

Older statistical models for evaluating latent variables forced the user to choose what type of distribution the latent variable should have. For example, Factor Analysis models assume continuous and normally distributed latent variables; Latent Class models assume discrete, multinomially distributed latent variables.

More recent models allow for flexible distributions, that combine latent factors and latent classes, which allows for greater flexibility for data exploration and hypothesis testing.

The talk introduces the concepts behind these more flexible, generalized latent variable models that can be used to explore the structure of latent variables.

## Introductory readings

- Bartholomew, Knott & Moustaki (2011).
*Latent Variable Models and Factor Analysis: A Unified Approach*(3rd Ed.). London: Wiley. - Muthen, B. O. (2002). Beyond SEM: General latent variable modelling.
*Behaviormetrika*, 29, 81-117.

## PDF slides

Download PDF slides of the presentation 'What is Latent Structure Analysis?'