bayesgmed_summary {BayesGmed}

Print a summary of the estimated causal mediation model

Description

Print a summary of the estimated causal mediation model

Usage

bayesgmed_summary(
  model = NULL,
  level = 0.95,
  pars = c("NDE_control", "NDE_treated", "NIE_control", "NIE_treated", "ANDE", "ANIE",
    "TE"),
  digits = 3
)

Arguments

model

A stanfit object obtained from bayesgmed().

level

The "confidence" level that defines the limits of the credible intervals (default is .95, i.e. 95% CIs).

pars

The parameters to summarize (defaults to main causal effect estimands similar to the mediation package; see Details).

digits

The number of decimal points to display in the output (default is 3).

Details

After estimating a model with bayesgmed(), use bayesgmed_summary(fit) to display the estimated results, where fit is an object containing the fitted model. By default, bayesgmed_summary() only displays a subset of the estimated parameters: - NDE_control: direct effect estimate when the exposure level is set to the control value. - NDE_treated: direct effect estimate when the exposure level is set to the treated value. - NIE_control: mediated effect estimate when the exposure level is set to the control value. - NIE_treated: mediated effect estimate when the exposure level is set to the treated value. - ANDE: average direct effect of X on Y. - ANIE: average indirect effect of X on Y. - TE: the total effect of A on Y. To display all estimated parameters where all chains merged, set pars = NULL. This will print all parameters defined in the model definitions, including the most important ones: - alphaZ[]: parameter estimate of the confounders (i.e., X -> Y) relationship, listed in the order they are specified in the covariates argument of bayesgmed(). alpha[1] is the intercept. - alphaM: parameter estimate of the M -> Y relationship. - alphaA: parameter estimate of the A -> Y relationship. - betaZ: parameter estimate of the confounders (i.e., X -> M) relationship, listed in the order they are specified in the covariates argument of bayesgmed(). beta[1] is the intercept

To learn more about the additional parameters, refer to the Stan code (cat(get_stancode(fit))).

Value

A data.frame summarizing the estimated causal mediation model, including the following columns:

• Parameter: the name of the parameter.

• Mean: the mean of the parameter's posterior distribution.

• Median: the median of the parameter's posterior distribution.

• SE: the standard deviation of the parameter's posterior distribution.

• ci_lwr: the lower limit of the credible interval.

• ci_upr: the upper limit of the credible interval.

• n_eff: the number of efficient samples.

• Rhat: a value of 1.00 suggests model convergence.

Author(s)

Belay B. Yimer [e-mail]