“…This lack is likely a result of the history of using specific BN software products that are designed for discrete variables or provide only a limited selection of continuous distribution types. Several popular BN softwares nowadays accommodate hybrid networks (e.g., Hugin, Analytica, AgenaRisk, bnlearn in R), but their analytical use is still limited compared to discrete models (see, however, Moe et al 2020, for example), and requires more statistical expertise from both the modeler and the end user of the model or its results. It is also noteworthy that although conditional probability tables of discrete variables are able to accommodate and express nonlinearities, step functions, and other nonparametric functional responses between the variables, the continuous variables are often modeled using linear functions or other simple parametric functional forms.…”