The basic idea underlying Latent Class Analysis (LCA) is that there are unobserved subgroups of cases in the data. These unobserved subgroups form the categories of a categorical latent variable. In a statistical model of data with such unobserved subgroups, some of the parameters of the model will differ across these subgroups (the categorical latent variable). LCA is a subset of structural equation models and shares similarities with factor analysis. In factor analysis, the unobserved latent variables are continuous, whereas in LCA they are categorical (discrete). Outside social sciences, LCA models are often referred to as finite mixture models.

For more methods resources see:
http://www.methods.manchester.ac.uk