During the POLPAN Seminar on March 26, Dr Michal Kotnarowski (Institute of Philosophy and Sociology, PAS, and Institute of Political Studies, PAS) presented a paper entitled ‘Latent Class Analysis and its possible application in research on social structure.’

The presentation will consist of two parts. In the first part, I will present the Latent Class Analysis (LCA) technique. It is an analytical tool that allows recovering a hidden structure of relations between variables on the basis of observable data. In this sense, the LCA is similar to popular factor analysis, except that the identified hidden variables are categorical, and there are no strong assumptions about the character of observable variables in LCA. LCA is also treated as a classification technique. For this reason, it can be compared to other classification techniques such as hierarchical clustering or K-means clustering. When compared to other classification techniques, the LCA has fewer assumptions about observable data. Moreover, the choice of classification solution (i.e. the number of clusters) is based on statistical measures in the LCA, while it is somewhat arbitrary in other techniques. In this part, I will present the mathematical foundations of the LCA, the properties of the method, and examples of its successful application.
In the second part of the speech, I will present some preliminary ideas on the use of the LCA in determining the social class of respondents. I will focus on the possibilities and the potential problems related to using LCA in determining the social class. My analyses will be based on the POLPAN and the Polish National Election Study datasets.