Studying Affiliation Networks Through Cluster CA and Blockmodeling

In this paper, we propose a new joint approach for analyzing affiliation (two-mode) networks by using factorial methods and blockmodeling. In our recent work, we pursued the joint use of a given factorial method, i.e., MCA/MFA, and a clustering method, namely blockmodeling, but through distinct steps. Here we look for a strategy that permits us to apply the two methods simultaneously. To this aim, we propose a method that allows us to group individuals and variables simultaneously and directly for binary matrices, namely cluster correspondence analysis (cluster CA). This method can be adopted when dealing with affiliation matrices having a binary structure. Hence, we look at the way network positions (clusters) can be incorporated in cluster CA to verify if cluster CA can properly represent specific network structures. We illustrate our proposal through an empirical application on an affiliation network of stage co-productions.