estimate_subgroup_ate: Find difference in ATE between subgroups
In ensley-nexant/cfeval: Causal Forest Evaluation and Visualization

Description Usage Arguments Value Examples

One heuristic for evaluating whether a causal forest has detected heterogeneity is to group observations into high- and low-CATE subgroups, estimate the ATE for each group, and determine whether the difference is significantly different than zero.

1	estimate_subgroup_ate(fit, subgroup, level = 0.9)

`fit`	A trained causal forest object from `causal_forest`
`subgroup`	A logical vector indicating whether an observation belongs to the subgrouping
`level`	The width of the reported confidence interval

A one-row tibble with the following columns:

in_est, in_se: The ATE estimate and corresponding standard error among observations for which subgroup = TRUE
out_est, out_se: The ATE estimate and corresponding standard error among observations for which subgroup = FALSE
diff_est: The difference in ATE: in_est - out_est
diff_ci_low, diff_ci_high: The confidence interval around diff_est

## Not run: 
 require(grf)

 n <- 2000; p <- 10

 X <- matrix(rnorm(n * p), n, p)
 W <- rbinom(n, 1, 0.4 + 0.2 * (X[, 1] > 0))
 Y <- pmax(X[, 1], 0) * W + X[, 2] + pmin(X[, 3], 0) + rnorm(n)

 cf <- causal_forest(X, Y, W)
 results <- tidy_cf(cf)

 # partition into high and low CATE subgroups
 estimate_subgroup_ate(cf, results$cate > median(results$cate))

## End(Not run)