Posterior distribution (5th, 25th, 50th, 75th, and 95th percentiles) of the weights |$ \boldsymbol{\alpha}_{\rm opt}$| for risk factors and the spatial penalty term when using linear utility function and data in year 2007–2012 (left), year 2007–2013 (middle), and year 2007–2014 (right) as training data, respectively. The risk factors include temperature (standardized), precipitation (standardized), current disease rate (logit), and current neighborhood disease rate (logit). (a) Training data: year 2007–2012. (b) Training data: year 2007–2013. (c) Training data: year 2007–2014.