Latent class analysis is a technique of assessing whether the set of observations including q categorical variables, in specific, binary variables, consists of the number of different groups or the classes within which the variables are independent. It is essentially a finite mixture model in which the component distributions are product of the q Bernoulli distributions, one for each of the binary variables in data. Parameters in such type of models can be estimated by the maximum likelihood estimation by means of the EM algorithm. It can be considered as either an analogue of the factor analysis for categorical variables, or the model of cluster analysis for such type of data.