placebo_test: Placebo permutation test for pensynth

In pensynth: Penalized Synthetic Control Estimation

Placebo permutation test for pensynth

Description

Perform a permutation test on a pensynth object, in the sense of Abadie, Diamond, and Hainmueller (2010). The pensynth method is performed multiple times, treating each donor as the treated unit and the treated unit with the remaining donors as the donor units.

Usage

placebo_test(object, Y1, Y0, verbose = TRUE)

## S3 method for class 'pensynth'
placebo_test(object, Y1, Y0, verbose = TRUE)

## S3 method for class 'cvpensynth'
placebo_test(object, Y1, Y0, verbose = TRUE)

Arguments

object	a fitted pensynth or cvpensynth object
Y1	the post-intervention outcome of the treated unit
Y0	the post-intervention outcome of the donor units
verbose	boolean whether to print progress messages. Default on if in an interactive session. (with N_donors columns)

Details

Note that this function updates the original call in order to re-estimate the synthetic control on the permuted data. Ensure that the data is available to the placebo test function (i.e., avoid complex environment functions such as with()), and ensure that the data does not change between estimating the original object and calling this function.

Value

A list with two elements

• E1, the treated unit effect(s), computed as Y1 - Y0 %*% w

• E0, the donor unit effects, computed in the same way but using the permutation test's weights.

• ATE1, the estimated ATE of the treated unit(s)

• ATE0, the estimated ATE of the donor units

References

Abadie, A., Diamond, A., & Hainmueller, J. (2010). Synthetic control methods for comparative case studies: Estimating the effect of California’s tobacco control program. Journal of the American statistical Association, 105(490), 493-505.

See Also

pensynth(), cv_pensynth(), plot.pensynthtest(), stats::update()

Examples

set.seed(45)

# simulate data with an effect of 0.8 SD
dat <- simulate_data_synth(treatment_effect = .8)

# fit a model
fit <- pensynth(dat$X1, dat$X0, lambda = 1e-5)

# Perform placebo test
test <- placebo_test(fit, dat$Y1, dat$Y0)
plot(test)
abline(h = .8, lty = 2)
legend("bottomright", lty = 2, legend = "true effect")

# compute a pseudo p-value based on ATE in
# the post-intervention time period
ref_dist <- stats::ecdf(test$ATE0)
1 - ref_dist(test$ATE1)

pensynth documentation built on May 7, 2026, 9:06 a.m.