Benchmarks for the strength of latent variables using observed covariates

Compute benchmarks for the strength of latent variables, under the assumption that the gains in explanatory power due to latent variables is proportional to the gains of observed covariates.

dml_benchmark(model, benchmark_covariates, dreg = NULL, yreg = NULL)

Arguments

model: an object of class dml.
benchmark_covariates: the observed covariates to benchmark. Either a character vector of column names in the model's x (each benchmarked on its own), or a named list where each element is a character vector of column names to drop together in the refit (e.g. the dummy columns of a factor: list(region = c("region3", "region4"))). List element names become the row labels; unnamed elements are labelled by the column name (singletons) or the columns joined by "+".
dreg, yreg: optional learner specifications (as in dml) for the refit, overriding those stored in model$call.

Value

An object of class dml_benchmark containing benchmark results.

Examples

data("pension")
set.seed(3); idx <- sample(nrow(pension), 2000)
y <- pension$net_tfa[idx]
d <- pension$e401[idx]
x <- model.matrix(~ -1 + age + inc + educ + fsize + marr + twoearn + pira + hown,
                  data = pension[idx, ])
fit <- dml(y, d, x, model = "plm", cf.folds = 2, cf.seed = 3, verbose = FALSE)

# observed gains of income, from a refit without it
bench <- dml_benchmark(fit, benchmark_covariates = "inc")
#> 
#> === Computing benchmarks using covariate: inc  ===
#> 
summary(bench)
#> Covariate benchmarks:
#> 
#>     gain.Y se.gain.Y gain.D se.gain.D     rho se.rho     delta se.delta
#> inc 0.1899    0.0687 0.0332    0.0104 -0.3338 0.1087 -3092.902 952.4735
#> 
#> Combined across cross-fitting repetitions using the median method.