A Shiny Application for End-to-End Bayesian Decision Network Analysis and Web-Deployment

A Shiny application for learning Bayesian Decision Networks from data. This package can be used for probabilistic reasoning (in the observational setting), causal inference (in the presence of interventions) and learning policy decisions (in Decision Network setting). Functionalities include end-to-end implementations for data-preprocessing, structure-learning, exact inference, approximate inference, extending the learned structure to Decision Networks and policy optimization using statistically rigorous methods such as bootstraps, resampling, ensemble-averaging and cross-validation. In addition to Bayesian Decision Networks, it also features correlation networks, community-detection, graph visualizations, graph exports and web-deployment of the learned models as Shiny dashboards.

Reference Analysis for Bayesian Meta-Analysis

Functionality for performing a principled reference analysis in the Bayesian normal-normal hierarchical model used for Bayesian meta-analysis, as described in Ott, Plummer and Roos (2021) . Computes a reference posterior, induced by a minimally informative improper reference prior for the between-study (heterogeneity) standard deviation. Determines additional proper anti-conservative (and conservative) prior benchmarks. Includes functions for reference analyses at both the posterior and the prior level, which, given the data, quantify the informativeness of a heterogeneity prior of interest relative to the minimally informative reference prior and the proper prior benchmarks. The functions operate on data sets which are compatible with the 'bayesmeta' package.

Sparse Generalized Additive Models

Fits sparse frequentist GAMs (SF-GAM) for continuous and discrete responses in the exponential dispersion family with the group lasso, group smoothly clipped absolute deviation (SCAD), and group minimax concave (MCP) penalties . Also fits sparse Bayesian generalized additive models (SB-GAM) with the spike-and-slab group lasso (SSGL) penalty of Bai et al. (2021) . B-spline basis functions are used to model the sparse additive functions. Stand-alone functions for group-regularized negative binomial regression, group-regularized gamma regression, and group-regularized regression in the exponential dispersion family with the SSGL penalty are also provided.

Bayesian Distributed Lag Non-Linear Models (B-DLNM)

A Bayesian framework for estimating distributed lag linear and non-linear models. Model fitting is implemented using Integrated Nested Laplace Approximation (R package 'INLA'), together with prediction and visualization of exposure-lag-response associations. Additional functions allow estimation of optimal exposure values (e.g., minimum mortality temperature) and computation of attributable fractions and numbers. Models with 'crossbasis' or 'onebasis' terms are supported (R package 'dlnm').

Bayesian Analysis of Finite Mixture of Plackett-Luce Models

Fit finite mixtures of Plackett-Luce models for partial top rankings/orderings within the Bayesian framework. It provides MAP point estimates via EM algorithm and posterior MCMC simulations via Gibbs Sampling. It also fits MLE as a special case of the noninformative Bayesian analysis with vague priors. In addition to inferential techniques, the package assists other fundamental phases of a model-based analysis for partial rankings/orderings, by including functions for data manipulation, simulation, descriptive summary, model selection and goodness-of-fit evaluation. Main references on the methods are Mollica and Tardella (2017) and Mollica and Tardella (2014) .

Censored Distributions for 'greta'

Provides additional censored distributions for use with 'greta', a probabilistic programming framework for Bayesian modeling. Includes censored versions of Normal, Log-Normal, Student's T, Gamma, Exponential, Weibull, Pareto, and Beta distributions with support for right, left, and interval censoring. For details on 'greta', see Golding (2019) . The methods are implemented using 'TensorFlow' and 'TensorFlow Probability' for efficient computation.

Stepwise Elimination and Term Reordering for Mixed-Effects Regression

Finds the largest possible regression model that will still converge for various types of regression analyses (including mixed models and generalized additive models) and then optionally performs stepwise elimination similar to the forward and backward effect-selection methods in SAS, based on the change in log-likelihood or its significance, Akaike's Information Criterion, the Bayesian Information Criterion, the explained deviance, or the F-test of the change in R².

Dirichlet Process Clustering with Dissimilarities

A Bayesian hierarchical model for clustering dissimilarity data using the Dirichlet process. The latent configuration of objects and the number of clusters are automatically inferred during the fitting process. The package supports multiple models which are available to detect clusters of various shapes and sizes using different covariance structures. Additional functions are included to ensure adequate model fits through prior and posterior predictive checks.

Kernel Density Estimation for Heaped and Rounded Data

In self-reported or anonymised data the user often encounters heaped data, i.e. data which are rounded (to a possibly different degree of coarseness). While this is mostly a minor problem in parametric density estimation the bias can be very large for non-parametric methods such as kernel density estimation. This package implements a partly Bayesian algorithm treating the true unknown values as additional parameters and estimates the rounding parameters to give a corrected kernel density estimate. It supports various standard bandwidth selection methods. Varying rounding probabilities (depending on the true value) and asymmetric rounding is estimable as well: Gross, M. and Rendtel, U. (2016) (). Additionally, bivariate non-parametric density estimation for rounded data, Gross, M. et al. (2016) (), as well as data aggregated on areas is supported.

Monte Carlo and Bayesian Stop-Signal Reaction Time Estimation

Estimates stop-signal reaction time (SSRT) in the stop-signal task using the integration and mean methods described by Verbruggen and colleagues (2019) . In addition to point estimates, the package provides Monte Carlo tools (nonparametric bootstrap confidence intervals, parametric ex-Gaussian simulation, minimum-trial-count and power analysis, and sensitivity analysis under violations of the horse-race assumptions) and Bayesian estimation via 'Stan', including single-subject and hierarchical ex-Gaussian horse-race models with an optional trigger-failure parameter following Matzke and colleagues (2013) , posterior inhibition functions, and posterior predictive checks. The Bayesian layer works with either the 'cmdstanr' or 'rstan' backend.