B0340
Title: A Hidden Markov model addressing ordinal response for non-decreasing processes
Authors: Carlos Javier Perez Sanchez - University of Extremadura (Spain)
Lizbeth Naranjo Albarran - Universidad Nacional Autonoma de Mexico UNAM (Mexico) [presenting]
Yolanda Campos-Roca - Universidad de Extremadura (Spain)
Abstract: Several investigations have recently considered the use of acoustic parameters extracted from speech recordings as an objective and non-invasive tool to perform diagnosis and monitoring of Parkinson's Disease (PD). Repeated speech recordings were obtained from which several acoustic characteristics were extracted. The objective is to monitor the progression of people with PD in the Hoehn and Yahr scale. A Hidden Markov Model (HMM) addressing the ordinal response with some missing data for monotonic non-decreasing processes is proposed. This model used the strength of the HMM to track the progression of the disease at the same time that handles ordinal response with missingness data and non-decreasing of the stages through the time. The way the model is defined allows the derivation of an efficient MCMC algorithm.
