Estimating a Causal Exposure Response Function with a Continuous Error-Prone Exposure: A Study of Fine Particulate Matter and All-Cause Mortality
erf.R Includes baseline functions for fitting an exposure response function (ERF) without measurement error correction. Code in this script is later used by bart-erf.R, bayes-erf.R, and spatial-erf.R.
bart-erf.R Includes the function bart_erf() which fits a measurement error corrected ERF using multiple imputation and a Bayesian additive regression tree (BART) outcome model.
bayes-erf.R An alternative function to bart_erf() which fits a generalized linear model of the outcome before regressing the pseudo-outcome onto the exposure support.
spatial-erf.R An alternative function to bart_erf() which incorporates a spatial autocorrelation random variable into the cluster-level exposure model.
auxiliary.R Additional functions used intermittently throughout the manuscript including a function to estimate the regression calibrated grid-level exposures, a function to compute the highest posterior density interval when the bart_erf() and bayes_erf() is used without smoothing, and a function to check the adjacency matrix for the spatial autocorrelation component.
/sim Code for running the numerical studies is contained in this directory. Also included in this folder is test.R for unit testing and gen-data.R which includes functions to generate simulated data and the true exposure response function.