Outcome models:

[azev77]

Broad classes of outcome models (partly based on Athey et al) :
note: time is the horizontal dimension, unit is the vertical dimension

1. Synthetic Control Methods or Vertical Regression (VR)
2. Unconfoundedness or Horizontal Regression (HZ)
3. Matrix Completion (MC) Methods: {using nuclear norm}
4. Ensemble Methods (Vertical Cross-Validation (VC) or Horizontal Cross-Validation (HC))
5. Time-series forecast methods {especially w/ prediction intervals}

• classic ADH (Convex, no intercept): Optim.jl vs Jump.jl + solver?
• Classic Diff-in-Diff (two-way FE) as a reference/starting point
• InteractiveFixedEffectModels.jl {basically GSynth.R}
• Any model from MLJ.jl & Ensemble
• Auto Arima: ARCHModels.jl
• https://github.com/google/CausalImpact
• Matrix Completion MCPanel.R
• https://github.com/synth-inference/synthdid