Sensitivity Assessment to Unmeasured Confounding with Multiple Treatments

A sensitivity analysis approach for unmeasured confounding in observational data with multiple treatments and a binary outcome. This approach derives the general bias formula and provides adjusted causal effect estimates in response to various assumptions about the degree of unmeasured confounding. Nested multiple imputation is embedded within the Bayesian framework to integrate uncertainty about the sensitivity parameters and sampling variability. Bayesian Additive Regression Model (BART) is used for outcome modeling. The causal estimands are the conditional average treatment effects (CATE) based on the risk difference. For more details, see paper: Hu L et al. (2020) A flexible sensitivity analysis approach for unmeasured confounding with multiple treatments and a binary outcome with application to SEER-Medicare lung cancer data .

User-Friendly 'shiny' App for Bayesian Species Distribution Models

A user-friendly 'shiny' application for Bayesian machine learning analysis of marine species distributions. GLOSSA (Global Ocean Species Spatio-temporal Analysis) uses Bayesian Additive Regression Trees (BART; Chipman, George, and McCulloch (2010) ) to model species distributions with intuitive workflows for data upload, processing, model fitting, and result visualization. It supports presence-absence and presence-only data (with pseudo-absence generation), spatial thinning, cross-validation, and scenario-based projections. GLOSSA is designed to facilitate ecological research by providing easy-to-use tools for analyzing and visualizing marine species distributions across different spatial and temporal scales. Optionally, pseudo-absences can be generated within the environmental space using the external package 'flexsdm' (not on CRAN), which can be downloaded from < https://github.com/sjevelazco/flexsdm>; this functionality is used conditionally when available and all core features work without it.

Basis Expansions for Regression Modeling

Provides various basis expansions for flexible regression modeling, including random Fourier features (Rahimi & Recht, 2007) < https://proceedings.neurips.cc/paper_files/paper/2007/file/013a006f03dbc5392effeb8f18fda755-Paper.pdf>, exact kernel / Gaussian process feature maps, Bayesian Additive Regression Trees (BART) (Chipman et al., 2010) prior features, and a helpful interface for n-way interactions. The provided functions may be used within any modeling formula, allowing the use of kernel methods and other basis expansions in modeling functions that do not otherwise support them. Along with the basis expansions, a number of kernel functions are also provided, which support kernel arithmetic to form new kernels. Basic ridge regression functionality is included as well.

Tools for Working with URLs and HTTP

Useful tools for working with HTTP organised by HTTP verbs (GET(), POST(), etc). Configuration functions make it easy to control additional request components (authenticate(), add_headers() and so on).

Create Visualisations for BART Models

Investigating and visualising Bayesian Additive Regression Tree (BART) (Chipman, H. A., George, E. I., & McCulloch, R. E. 2010) model fits. We construct conventional plots to analyze a model’s performance and stability as well as create new tree-based plots to analyze variable importance, interaction, and tree structure. We employ Value Suppressing Uncertainty Palettes (VSUP) to construct heatmaps that display variable importance and interactions jointly using colour scale to represent posterior uncertainty. Our visualisations are designed to work with the most popular BART R packages available, namely 'BART' Rodney Sparapani and Charles Spanbauer and Robert McCulloch 2021 , 'dbarts' (Vincent Dorie 2023) < https://CRAN.R-project.org/package=dbarts>, and 'bartMachine' (Adam Kapelner and Justin Bleich 2016) .

Covariance Inference and Decompositions for Tensor Datasets

A collection of functions for Kronecker structured covariance estimation and testing under the array normal model. For estimation, maximum likelihood and Bayesian equivariant estimation procedures are implemented. For testing, a likelihood ratio testing procedure is available. This package also contains additional functions for manipulating and decomposing tensor data sets. This work was partially supported by NSF grant DMS-1505136. Details of the methods are described in Gerard and Hoff (2015) and Gerard and Hoff (2016) .

Visualization of BART and BARP using SHAP

Complex machine learning models are often difficult to interpret. Shapley values serve as a powerful tool to understand and explain why a model makes a particular prediction. This package computes variable contributions using permutation-based Shapley values for Bayesian Additive Regression Trees (BART) and its extension with Post-Stratification (BARP). The permutation-based SHAP method proposed by Strumbel and Kononenko (2014) is grounded in data obtained via MCMC sampling. Similar to the BART model introduced by Chipman, George, and McCulloch (2010) , this package leverages Bayesian posterior samples generated during model estimation, allowing variable contributions to be computed without requiring additional sampling. The BART model is designed to work with the following R packages: 'BART' , 'bartMachine' , and 'dbarts' < https://CRAN.R-project.org/package=dbarts>. For XGBoost and baseline adjustments, the approach by Lundberg et al. (2020) is also considered. The BARP model proposed by Bisbee (2019) was implemented with reference to < https://github.com/jbisbee1/BARP> and is designed to work with modified functions based on that implementation. BARP extends post-stratification by computing variable contributions within each stratum defined by stratifying variables. The resulting Shapley values are visualized through both global and local explanation methods.

A Simple and Robust JSON Parser and Generator for R

A reasonably fast JSON parser and generator, optimized for statistical data and the web. Offers simple, flexible tools for working with JSON in R, and is particularly powerful for building pipelines and interacting with a web API. The implementation is based on the mapping described in the vignette (Ooms, 2014). In addition to converting JSON data from/to R objects, 'jsonlite' contains functions to stream, validate, and prettify JSON data. The unit tests included with the package verify that all edge cases are encoded and decoded consistently for use with dynamic data in systems and applications.

Tools for Spell Checking in R

Spell checking common document formats including latex, markdown, manual pages, and description files. Includes utilities to automate checking of documentation and vignettes as a unit test during 'R CMD check'. Both British and American English are supported out of the box and other languages can be added. In addition, packages may define a 'wordlist' to allow custom terminology without having to abuse punctuation.

Bayesian Models for Dissolution Testing

Fits Bayesian models (amongst others) to dissolution data sets that can be used for dissolution testing. The package was originally constructed to include only the Bayesian models outlined in Pourmohamad et al. (2022) . However, additional Bayesian and non-Bayesian models (based on bootstrapping and generalized pivotal quanties) have also been added. More models may be added over time.