A Bayesian Semi-Parametric Mixed Beta Regression Model for Disease Severity in Plants

Severity progress curves are used in plant disease epidemiology to describe temporal changes in the proportion of plant material compromised by the disease. For diseases with leaf symptoms, typically the damage of several leaves is assessed on each leaf on a particular scale and then averaged to a severity index (SI). The SI is often expressed in a 0-1 scale, which naturally leads to a beta distribution. In this paper we propose a Bayesian semiparametric beta regression to model the progress of disease severity. The model incorporates splines to estimate the population-average and plant-specific curves; additional terms related to the experiment design can be also included. One of the advantages of the proposed model is that it facilitates the comparison of curves between treatments across time. We applied the proposed model to Black Sigatoka disease on banana crop data from Isabela, Puerto Rico. The MCMC scheme of the proposed model was implemented in JAGS via the R2jags package. The interpretation of the analyses and the implications for the management of this disease are presented and discussed.