GitHub

TASC.jl

TASC.jl implements Time-Aware Synthetic Control for panel data with temporal dependence. It fits a low-rank linear Gaussian state-space model on the pre-treatment panel, then treats post-treatment treated outcomes as missing and uses donor outcomes to infer the untreated counterfactual path.

The package includes:

  • fit_tasc and predict_counterfactual for TASC estimation.
  • Model-based pointwise uncertainty from the smoothed latent-state covariance.
  • Multiple treated unit support through treated_rows.
  • A lightweight RecipesBase plotting recipe through tasc_plot.
  • Classical synthetic-control baselines and matrix preprocessing utilities ported from the Python package.
  • Synthetic data generators for simulation studies.

Installation

using Pkg
Pkg.add(url = "https://github.com/xiangao/TASC.jl")

For local development:

using Pkg
Pkg.develop(path = "/home/xao/projects/software/TASC.jl")

Quick Start

using TASC
using Statistics
using Random

Random.seed!(123)

Y, params, signal = simulate_tasc(N = 16, T = 60, d = 3, seed = 123)
T0 = 35

model = fit_tasc(Y; d = 3, T0 = T0, n_em = 20, tol = 1e-4)
pred = predict_counterfactual(model, Y)

att = mean(pred.effect[(T0 + 1):end])
ci_lower = pred.target .- 1.96 .* sqrt.(max.(pred.variance, 0.0))
ci_upper = pred.target .+ 1.96 .* sqrt.(max.(pred.variance, 0.0))

round(att, digits = 3)
-0.055

See the vignettes for a complete workflow.