Estimates causal contrasts using generalized linear models for different types of treatment effects (ATE, ATT) and outcomes (RR, RD). Supports handling multiply imputed datasets and allows flexibility in model specification.
Usage
causal_contrast_engine(
df,
Y,
X,
baseline_vars,
treat_0,
treat_1,
estimand = c("ATE", "ATT"),
type = c("RR", "RD"),
nsims = 200,
cores = parallel::detectCores(),
family = "gaussian",
weights = TRUE,
continuous_X = FALSE,
splines = FALSE,
vcov = "HC2",
verbose = FALSE
)Arguments
- df
Data frame or `mids` object containing the data.
- Y
Response variable name as a string.
- X
Treatment or exposure variable name as a string.
- baseline_vars
Vector of baseline covariate names.
- treat_0
Reference level of the treatment variable.
- treat_1
Treatment level of interest for comparison.
- estimand
Type of causal estimand ("ATE", "ATT"); defaults to both.
- type
Type of effect size ("RR" for Risk Ratio, "RD" for Risk Difference); defaults to both.
- nsims
Number of simulations for bootstrap; defaults to 200.
- cores
Number of cores for parallel processing; uses all available cores by default.
- family
Model family as a string or family object; defaults to "gaussian".
- weights
The name of the weights variable in the data frame, or NULL if no weights are to be used.
- continuous_X
Whether X is a continuous variable; defaults to FALSE.
- splines
Whether to apply spline transformation to X; defaults to FALSE.
- vcov
Type of variance-covariance matrix for standard error estimation; defaults to "HC2".
- verbose
Whether to print detailed output; defaults to FALSE.