Plot a regression tree predicting variables with lack of overlap — plot_predicted_common

Identify variables that predict lack of overlap

Usage

plot_predicted_common_support(
  .model,
  max_depth = 3,
  rule = c("both", "sd", "chi")
)

Arguments

.model: a model produced by `bartCause::bartc()`
max_depth: a number indicatin the max depth of the tree. Higher numbers are more prone to overfitting.
rule: one of c('both', 'sd', 'chi') denoting which rule to use to identify lack of support

Value

ggplot object

Author

George Perrett, Joseph Marlo

Examples

# \donttest{
data(lalonde)
confounders <- c('age', 'educ', 'black', 'hisp', 'married', 'nodegr')
model_results <- bartCause::bartc(
 response = lalonde[['re78']],
 treatment = lalonde[['treat']],
 confounders = as.matrix(lalonde[, confounders]),
 estimand = 'ate',
 commonSuprule = 'none'
)
#> fitting treatment model via method 'bart'
#> fitting response model via method 'bart'
plot_predicted_common_support (model_results)
#> no cases were removed under the standard deviation or chi-squared rules
#> NULL
plot_predicted_common_support (model_results, max_depth = 2, rule = 'chi')
#> no cases were removed under the chi squared rule
#> NULL
# }