Determinantal methods in latent class analysis

Some extensions of the existing determinantal methods for solving the accounting equations in latent class analysis are presented. These extensions cover more cases than previous methods, give rise to new sufficient conditions for identifiability of the latent class model, and give insight into the necessity of various sufficient conditions for identifiability. These implications to the identifiability problem are discussed.This work, part of a doctoral dissertation submitted to the Department of Statistics of the University of Chicago, was made possible by The RAND Corporation and by a grant to the Department of Statistics of the University of Chicago by the Rockefeller Foundation. I gratefully acknowledge the comments of T. W. Anderson, W. H. Kruskal, and D. L. Wallace and especially the comments of L. A. Goodman on this work.