mmb: Arbitrary Dependency Mixed Multivariate Bayesian Models

Supports Bayesian models with full and partial (hence arbitrary) dependencies between random variables. Discrete and continuous variables are supported, and conditional joint probabilities and probability densities are estimated using Kernel Density Estimation (KDE). The full general form, which implements an extension to Bayes' theorem, as well as the simple form, which is just a Bayesian network, both support regression through segmentation and KDE and estimation of probability or relative likelihood of discrete or continuous target random variables. This package also provides true statistical distance measures based on Bayesian models. Furthermore, these measures can be facilitated on neighborhood searches, and to estimate the similarity and distance between data points. Related work is by Bayes (1763) <doi:10.1098/rstl.1763.0053> and by Scutari (2010) <doi:10.18637/jss.v035.i03>.

install.packages("mmb")

Try the mmb package in your browser