linearDML

Overview

linearDML implements double/debiased machine learning (DML) Chernozhukov et al. (2018) for partially linear models and adds a method for flexibly specifying interactions with treatment variables.

Usage

dml.lm is the workhorse function of linearDML. Along with first stage DML predictions, dml.lm returns an lm object containing estimates from the second stage.

# dml with RF predictions

model.dml = dml.lm(iris
  , y_var = 'Sepal.Length' #outcome
  , x_vars = c('Sepal.Width', 'Petal.Width') #co-variates
  , d_vars = 'Petal.Length' #treatment
  , first_stage_family = 'rf')

summary(model.dml$model)

When using dml.lm to specify variable interactions, the h_vars argument expects a dataframe with three columns

• d - the name of the treatment variable to interact
• fx - the name of the covariate to interact
• fxd.name - the name of a calculated columns that defines the variable interaction

# Using h_vars, with covariate interaction terms
#
iris$sep.wid.2 = with(iris, Sepal.Width ^ 2)
iris$pet.len.sep.wid.2 = with(iris, Petal.Length * sep.wid.2)
iris$const = 1


h_vars = data.frame(d = c('Petal.Length', 'Petal.Length', 'Petal.Width'),
                    fx = c('const', 'sep.wid.2', 'const'),
                    fxd.name = c('Petal.Length', 'pet.len.sep.wid.2', 'Petal.Width'))

model.dml.h = dml.lm(iris
  , y_var = 'Sepal.Length'
  , x_vars = c('Sepal.Width')
  , h_vars = h_vars
  , first_stage_family = 'ols')

summary(model.dml.h$model)

Installation

You can install the latest version of linearDML from Github using the following command

devtools::install_github('ripl-org/linearDML')

References

Chernozhukov, V., Chetverikov, D., Demirer, M., Duflo, E., Hansen, C., Newey, W. and Robins, J. (2018), Double/debiased machine learning for treatment and structural parameters. The Econometrics Journal, 21: C1-C68. doi:10.1111/ectj.12097.

License

LICENSE

Copyright 2023 Innovative Policy Lab d/b/a Research Improving People's Lives ("RIPL"), Providence, RI. All Rights Reserved.