Skip to contents

Bootstrap posterior for a 2SLS effect as an edaphos_posterior

Source: R/causal_iv.R
causal_iv_posterior.Rd

Wraps causal_iv_fit_2sls() in a nonparametric bootstrap over rows (or over clusters if cluster is supplied) and packages the resulting vector of effect estimates as an edaphos_posterior() so that uncertainty_calibrate() and ggplot2::autoplot() apply uniformly.

Usage

causal_iv_posterior(
  data,
  exposure,
  outcome,
  instruments,
  covariates = NULL,
  B = 500L,
  cluster = NULL,
  seed = NULL,
  units = NULL
)

Arguments

data

A data frame.

exposure

Character; name of the endogenous exposure column.

outcome

Character; name of the outcome column.

instruments

Character vector; names of instrument columns. More instruments than exposures gives an over-identified model on which the Sargan test is applicable.

covariates

Optional character vector of exogenous-control column names included in both first and second stage.

B

Integer; number of bootstrap resamples. Default 500L.

cluster

Optional character; column in data to resample by block (e.g. "kmeans_cluster" for spatial resampling).

seed

Optional RNG seed.

units

Optional units string.

Value

An edaphos_posterior with query_type = "effect" and method = "bootstrap".