Bayesian probabilistic Parkinson’s epidemiology model
figureposted on 12.09.2017, 21:48 by Michael MacAskillMichael MacAskill, Toni L. Pitcher, Daniel J. Myall
Researchers at the New Zealand Brain Research Institute and University of Otago applied this Bayesian hierarchical Bernoulli regression model to estimate for the first time the prevalence and incidence of Parkinson’s disease in Māori and Pasifika within New Zealand. This figure was created to depict the modelling process, which began by identifying potential cases from a national database on the basis of medication use. The model was then trained on subsets of patients with independent diagnostic information from other sources, to assign a probability of Parkinson's to each case, contingent on age, sex, and ethnicity. The estimated totals were formed not by counting specific individuals but by summing the probability values of all cases.