Estimator of the local average treatment effect.

## Usage

```
ddml_late(
y,
D,
Z,
X,
learners,
learners_DXZ = learners,
learners_ZX = learners,
sample_folds = 2,
ensemble_type = "nnls",
shortstack = FALSE,
cv_folds = 5,
subsamples_Z0 = NULL,
subsamples_Z1 = NULL,
cv_subsamples_list_Z0 = NULL,
cv_subsamples_list_Z1 = NULL,
silent = FALSE
)
```

## Arguments

- y
The outcome variable.

- D
The binary endogenous variable of interest.

- Z
Binary instrumental variable.

- X
A (sparse) matrix of control variables.

- learners
May take one of two forms, depending on whether a single learner or stacking with multiple learners is used for estimation of the conditional expectation functions. If a single learner is used,

`learners`

is a list with two named elements:`what`

The base learner function. The function must be such that it predicts a named input`y`

using a named input`X`

.`args`

Optional arguments to be passed to`what`

.

If stacking with multiple learners is used,

`learners`

is a list of lists, each containing four named elements:`fun`

The base learner function. The function must be such that it predicts a named input`y`

using a named input`X`

.`args`

Optional arguments to be passed to`fun`

.`assign_X`

An optional vector of column indices corresponding to control variables in`X`

that are passed to the base learner.`assign_Z`

An optional vector of column indices corresponding to instruments in`Z`

that are passed to the base learner.

Omission of the

`args`

element results in default arguments being used in`fun`

. Omission of`assign_X`

(and/or`assign_Z`

) results in inclusion of all variables in`X`

(and/or`Z`

).- learners_DXZ
Optional argument to allow for different estimators of \(E[D \vert X, Z]\). Setup is identical to

`learners`

.- learners_ZX
Optional argument to allow for different estimators of \(E[Z\vert X]\). Setup is identical to

`learners`

.- sample_folds
Number of cross-fitting folds.

- ensemble_type
Ensemble method to combine base learners into final estimate of the conditional expectation functions. Possible values are:

`"nnls"`

Non-negative least squares.`"nnls1"`

Non-negative least squares with the constraint that all weights sum to one.`"singlebest"`

Select base learner with minimum MSPE.`"ols"`

Ordinary least squares.`"average"`

Simple average over base learners.

Multiple ensemble types may be passed as a vector of strings.

- shortstack
Boolean to use short-stacking.

- cv_folds
Number of folds used for cross-validation in ensemble construction.

- subsamples_Z0, subsamples_Z1
List of vectors with sample indices for cross-fitting, corresponding to observations with \(Z=0\) and \(Z=1\), respectively.

- cv_subsamples_list_Z0, cv_subsamples_list_Z1
List of lists, each corresponding to a subsample containing vectors with subsample indices for cross-validation. Arguments are separated for observations with \(Z=0\) and \(Z=1\), respectively.

- silent
Boolean to silence estimation updates.

## Value

`ddml_late`

returns an object of S3 class
`ddml_late`

. An object of class `ddml_late`

is a list
containing the following components:

`late`

A vector with the average treatment effect estimates.

`weights`

A list of matrices, providing the weight assigned to each base learner (in chronological order) by the ensemble procedure.

`mspe`

A list of matrices, providing the MSPE of each base learner (in chronological order) computed by the cross-validation step in the ensemble construction.

`psi_a`

,`psi_b`

Matrices needed for the computation of scores. Used in

`summary.ddml_late()`

.`learners`

,`learners_DXZ`

,`learners_ZX`

,`subsamples_Z0`

,`subsamples_Z1`

,`cv_subsamples_list_Z0`

,`cv_subsamples_list_Z1`

,`ensemble_type`

Pass-through of selected user-provided arguments. See above.

## Details

`ddml_late`

provides a double/debiased machine learning
estimator for the local average treatment effect in the interactive model
given by

\(Y = g_0(D, X) + U,\)

where \((Y, D, X, Z, U)\) is a random vector such that \(\operatorname{supp} D = \operatorname{supp} Z = \{0,1\}\), \(E[U\vert X, Z] = 0\), \(E[Var(E[D\vert X, Z]\vert X)] \neq 0\), \(\Pr(Z=1\vert X) \in (0, 1)\) with probability 1, \(p_0(1, X) \geq p_0(0, X)\) with probability 1 where \(p_0(Z, X) \equiv \Pr(D=1\vert Z, X)\), and \(g_0\) is an unknown nuisance function.

In this model, the local average treatment effect is defined as

\(\theta_0^{\textrm{LATE}} \equiv E[g_0(1, X) - g_0(0, X)\vert p_0(1, X) > p(0, X)]\).

## References

Ahrens A, Hansen C B, Schaffer M E, Wiemann T (2023). "ddml: Double/debiased machine learning in Stata." https://arxiv.org/abs/2301.09397

Chernozhukov V, Chetverikov D, Demirer M, Duflo E, Hansen C B, Newey W, Robins J (2018). "Double/debiased machine learning for treatment and structural parameters." The Econometrics Journal, 21(1), C1-C68.

Imbens G, Angrist J (1004). "Identification and Estimation of Local Average Treatment Effects." Econometrica, 62(2), 467-475.

Wolpert D H (1992). "Stacked generalization." Neural Networks, 5(2), 241-259.

## See also

Other ddml:
`ddml_ate()`

,
`ddml_fpliv()`

,
`ddml_pliv()`

,
`ddml_plm()`

## Examples

```
# Construct variables from the included Angrist & Evans (1998) data
y = AE98[, "worked"]
D = AE98[, "morekids"]
Z = AE98[, "samesex"]
X = AE98[, c("age","agefst","black","hisp","othrace","educ")]
# Estimate the local average treatment effect using a single base learner,
# ridge.
late_fit <- ddml_late(y, D, Z, X,
learners = list(what = mdl_glmnet,
args = list(alpha = 0)),
sample_folds = 2,
silent = TRUE)
summary(late_fit)
#> LATE estimation results:
#>
#> Estimate Std. Error t value Pr(>|t|)
#> -0.2675741 0.1978461 -1.352436 0.176236
# Estimate the local average treatment effect using short-stacking with base
# learners ols, lasso, and ridge.
late_fit <- ddml_late(y, D, Z, X,
learners = list(list(fun = ols),
list(fun = mdl_glmnet),
list(fun = mdl_glmnet,
args = list(alpha = 0))),
ensemble_type = 'nnls',
shortstack = TRUE,
sample_folds = 2,
silent = TRUE)
summary(late_fit)
#> LATE estimation results:
#>
#> Estimate Std. Error t value Pr(>|t|)
#> nnls -0.1831584 0.1950704 -0.9389348 0.3477642
```