LETTER Communicated by Pearl Chiu Infinite Continuous Feature Model for Psychiatric Comorbidity Analysis Isabel Valera ivalera@mpi-sws.org Max Planck Institute for Software Systems, 67663 Kaiserslautern, Germany Francisco J. R. Ruiz franrruiz@columbia.edu Department of Signal Processing and Communications, University Carlos III in Madrid, 28911 Leganes, Spain; Gregorio Mara ˜ n´ on Health Research Institute, 28007 Madrid, Spain; and Department of Computer Science, Columbia University, New York, NY 10027, U.S.A. Pablo M. Olmos olmos@tsc.uc3m.es Department of Signal Processing and Communications, University Carlos III in Madrid, 28911 Leganes, Madrid, and Gregorio Mara ˜ n´ on Health Research Institute, 28007 Madrid, Spain Carlos Blanco cblanco@nyspi.columbia.edu Department of Psychiatry, New York State Psychiatric Institute, Columbia University, New York, NY 10032, U.S.A. Fernando Perez-Cruz Fernando.Perez-Cruz@Alcatel-Lucent.com Department of Signal Processing and Communications, University Carlos III in Madrid, 28911 Leganes, Madrid; Gregorio Mara ˜ n´ on Health Research Institute, 28007 Madrid, Spain; and Bell Labs, Alcatel-Lucent, New Providence, NJ 07974, U. S. A. We aim at finding the comorbidity patterns of substance abuse, mood and personality disorders using the diagnoses from the National Epidemio- logic Survey on Alcohol and Related Conditions database. To this end, we propose a novel Bayesian nonparametric latent feature model for cat- egorical observations, based on the Indian buffet process, in which the latent variables can take values between 0 and 1. The proposed model has several interesting features for modeling psychiatric disorders. First, I. Valera and F. J. R. Ruiz contributed equally to this letter. Neural Computation 28, 354–381 (2016) c  2016 Massachusetts Institute of Technology doi:10.1162/NECO_a_00805