Mixed mode data clustering: an approach based on tectrachoric correlations

In this paper we face the problem of clustering mixedmode data by assuming that the observed binary variables aregenerated from latent continuous variables. We perform a principalcomponents analysis on the matrix of tetrachoric correlations and wethen estimate the scores of each latent variable and construct adata matrix with continuous variables to be used in fully Guassianmixture models or in the k-means cluster analysis. The calculationof the expected a posteriori (EAP) estimates may proceed by simplyconsidering a limited number of quadrature points. Results on asimulation study and on a real data set are reported.