LERCA: Local Exposure-Response Confounding Adjustment.
In gpapadog/LERCA: Local Exposure-Response Confounding Adjustment

Description Usage Arguments

Function that calculates the mean exposure response curve allowing for differential confounding at different exposure levels.

LERCA(dta, chains, Nsims, K, cov_cols, omega = 5000, mu_priorX = NULL,
  Sigma_priorX = NULL, mu_priorY = NULL, Sigma_priorY = NULL,
  alpha_priorX = 0.001, beta_priorX = 0.001, alpha_priorY = 0.001,
  beta_priorY = 0.001, starting_cutoffs = NULL,
  starting_alphas = NULL, starting_coefs = NULL,
  starting_vars = NULL, approx_likelihood = TRUE,
  prop_distribution = c("Uniform", "Normal"), normal_percent = 1,
  plot_every = 0, comb_probs = c(0.01, 0.5, 0.99),
  split_probs = c(0.2, 0.95), s_upd_probs = c(0.8, 0.1, 0.1),
  alpha_probs = c(0.01, 0.5, 0.99), min_exper_sample = 20,
  jump_slope_tune = 0.05)

`dta`	A data set including a column of the exposure of interest as X, the outcome of interest as Y, and all potential confounders as C1, C2, ...
`chains`	The number of MCMC chains.
`Nsims`	The number of posterior samples per chain.
`K`	The number of points in the experiment configuration.
`cov_cols`	The indices of the columns in dta corresponding to the potential confounders.
`omega`	The parameter of the BAC prior on the inclusion indicators.
`mu_priorX`	The mean of the normal prior on the coefficients of the exposure model. Numeric vector of length equal to the number of potential confounders + 1 with the first entry corresponding to the intercept. If left NULL, it is set to 0 for all parameters.
`Sigma_priorX`	Covariance matrix of the normal prior on the regression coefficients of the exposure model. If left NULL, it is set to diagonal with entries 100 ^ 2 for all parameters.
`mu_priorY`	The mean of the normal prior on the coefficients of the outcome model. Numeric vector with entries corresponding to intercept, slope of exposure, and potential covariates. If left NULL, it is set to 0 for all parameters.
`Sigma_priorY`	The covariance matrix of the normal prior on the regression coefficients of the outcome model. If left NULL, it is set to diagonal with entries 100 ^ 2 for all parameters.
`alpha_priorX`	The shape parameter of the inverse gamma prior on the residual variance of the exposure model.
`beta_priorX`	The rate parameter of the inverse gamma prior on the residual variance of the exposure model.
`alpha_priorY`	The shape parameter of the inverse gamma prior on the residual variance of the outcome model.
`beta_priorY`	The rate parameter of the inverse gamma prior on the residual variance of the outcome model.
`starting_alphas`	Array with dimensions corresponding to the model (exposure / outcome), the experiment, and the potential confounders. Entries 0/1 represent exclusion/inclusion of the covariate in the corresponding model.
`starting_coefs`	Array with the starting values of all coefficients. Dimensions are: Exposure/Outcome model, chains, experiments, and covariate (intercept, coefficient of exposure, covariates). The coefficient of exposure should be NA for the exposure model.
`starting_vars`	Array including the starting values for the residual variances. Dimensions correspond to: Exposure/Outcome model, chains, and experiment.
`approx_likelihood`	Logical. If set to TRUE the likelihood of the data in the jump over and jump within moves will be calculated based on the BIC approximation. Defaults to TRUE. Option FALSE not supported for now.
`prop_distribution`	Character vector. Options include 'Uniform' or 'Normal' representing the type of distribution that will be used to propose a move of the cutoffs in the separate update. Defaults to uniform.
`normal_percent`	Numeric. Parameter controling the width of a normal proposal for the experiment configuration. Used only when prop_distribution is set to Normal. Smaller values represent smaller variance of the truncated normal proposal distribution.
`plot_every`	Integer. Plot the locations of the experiment configuration every plot_every iteration. Defaults to 0 leading to no plotting.
`comb_probs`	When two experiments are combined, comb_probs represents the probability of alpha = 1 when 0, 1, and 2 corresponding alphas are equal to 1. Vector of length 3. Defaults to (0.01, 0.5, 0.99).
`split_probs`	When one experiment is split, split_probs describes the probability that the alpha of a new experiment is equal to 1, when the alpha of the current experiment is 0, and when it is 1. Vector of length 2. Defaults to (0.2, 0.95).
`s_upd_probs`	Numeric of length three. The probability that each of the separate, jump over, and jump within moves is proposed. Defaults to (0.8, 0.1, 0.1).
`alpha_probs`	The probability that a proposed alpha is equal to 1, when 0, 1, and 2 alphas of the surrounding experiments are equal to 1. Vector of length 3. Defaults to (0.01, 0.5, 0.99).
`min_exper_sample`	The minimum number of observations within an experiment. Defaults to 20.
`jump_slope_tune`	The standard deviation of the proposal on the slopes for the jump over move. Defaults to 0.05.
`starting`	cutoffs Matrix with rows corresponding to different chains. Each row includes K ordered values of MCMC starting cutoffs. If left NULL, random started values are used.