ci_wald: Confidence Intervals for Conditional Effects and Evaluation...
In Netflix/sherlock: Causal Machine Learning for Segment Discovery and Analysis

Wald-style confidence intervals for data.tables with additional columns for estimates of the conditional average treatment effect (CATE) and related segmentation evaluation measures.

1
2
3

ci_wald(estimation_results, segment_by = NULL, param_type = c("cate",
  "effect_measure"), outcome_type = c("continuous", "binary"),
  coverage = getOption("sherlock.ci_covers"))

`estimation_results`	A `data.table` containing the input data, augmented to include cross-validated nuisance parameter estimates, an estimate of the CATE, and a treatment rule assigned based on the estimated CATE via `assign_rule`. This input object should be generated by successive calls to `set_est_data` and `est_cate`, or through a wrapper function that composes these function calls automatically. For cases in which `assign_rule` has been called prior, any additional information is preserved.
`segment_by`	A `character` vector specifying the column names in `data_obs` that correspond to the covariates over which segmentation should be performed. This should be a strict subset of `baseline`.
`param_type`	A `character` string (of length one) indicating the type of estimates for which a Wald-style confidence interval is to be computed. The choice of `"cate"` indicates that this routine is being invoked thru `summarize_segments` while `"effect_measure"` indicates that the input was generated by `est_effect`.
`outcome_type`	A `character` string (of length one) indicating whether the outcome variable is binary or continuous-valued. For cases in which the outcome is binary, the confidence interval for the parameter estimate is constructed on the logit scale and then back-transformed, in order to respect the fact that the estimate must lie in the unit interval.
`coverage`	A `numeric` indicating the nominal rate at which the Wald-style confidence interval is intended to cover the target parameter.

A data.table matching exactly the input provided in estimation_results, augmented to include the lower and upper bounds from the Wald-style confidence intervals constructed.

Netflix/sherlock documentation built on Dec. 17, 2021, 5:22 a.m.