About the Package
The R package causalbkmr consists of three parts: BKMR-CMA for Bayesian Kernel Machine Regression-Causal Mediation Analysis, MI-BKMR for Multiple Imputation BKMR, and g-BKMR for Bayesian Kernel Machine Regression for time-varying exposures and time-varying confounders.
We welcome your feedback and questions (email zc2326@columbia.edu)!
Installation
You can install the latest version of causalbkmr via:
devtools::install_github("zc2326/causalbkmr")Load causalbkmr:
BKMR-CMA
A command that implements the BKMR-CMA method, wich allows the estimation of direct and indirect health effects of multiple environmental exposures through a single mediator.
We use BKMR for the mediator and outcome regression models since BKMR allows for all possible nonlinearities and interactions among the elements included in the kernel with the specified outcome, without a prior specification, and credible intervals a BKMR fit inherently control for multiple testing due to the Bayesian nature of the model and the prior specification.
We predict counterfactuals using the posterior predictive distributions of the mediator and the outcome and present an algorithm for estimation of mediation effects. We also conduct a simulation study to compare how our approach performs relative to current mediation methods that assume a restrictive linear relationship between the exposure, mediator, and outcome.
Cite the paper: Bayesian kernel machine regression-causal mediation analysis
See the Quick Start guide for BKMR-CMA. See the package website for an overview of statistical modeling approaches.
BKMR-MI
A command that is used for valid estimation of environmental mixture effects and evaluation of uncertainty in the presence of missing data, which are imputed using multiple imputation techniques. The commands combine information from multiple Bayesian kernel machine regression (BKMR) models fit using the bkmr R package (Bobb et al. 2015, Valeri et al. 2017, Bobb et al. 2018, Anglen Bauer et al. 2019). The commands also produce effective visualizations of the estimated causal effects and dose-response relationships. The package contains functions to be used with BKMR MI fits to create a data frame for plotting with ggplot.
The data BKMRfits10 is simulated data with 10 BKMRfits, each is a BKMR fit using multiple imputed data with size n = 500.
See the Quick Start guide for BKMR-CMA. See the package website for an overview of statistical modeling approaches.
g-BKMR
Exposure to environmental chemicals has been shown to rewire development affecting later health status. Quantifying the joint effect of environmental mixtures over time is crucial to determine intervention timing. However, causal interpretation of longitudinal environmental mixture studies encounters challenges. There is no statistical approach that allows simultaneously for time-varying confounding, flexible modeling, and variable selection when examining the effect of multiple, correlated, and time-varying exposures. To address these gaps, we develop a causal inference method, g-BKMR, which enables us to estimate nonlinear, non-additive effects of time-varying exposures and time-varying confounders, while also allowing for variable selection.
