mlsynth

https://mlsynth.readthedocs.io
Admin
mlsynth is an open-source Python toolbox of synthetic-control methods for program evaluation.
Tokens:5,033
Snippets:30
Trust Score:9.6
Update:1 month ago
Show doc for...
Tokens:
Raw
### Install mlsynth from GitHub

Source: https://mlsynth.readthedocs.io

Use pip to install the latest version of the library directly from the GitHub repository.

```bash
pip install -U git+https://github.com/jgreathouse9/mlsynth.git
```

--------------------------------

### Install mlsynth with optional extras

Source: https://mlsynth.readthedocs.io

Install specific optional backends using bracket syntax. Ensure the specifier is quoted to prevent shell globbing.

```bash
# SCIP solver for SYNDES / MAREX
pip install -U "mlsynth[design] @ git+https://github.com/jgreathouse9/mlsynth.git"

# NumPyro for SPOTSYNTH's Bayesian mode
pip install -U "mlsynth[bayes] @ git+https://github.com/jgreathouse9/mlsynth.git"

# everything
pip install -U "mlsynth[all] @ git+https://github.com/jgreathouse9/mlsynth.git"
```

--------------------------------

### Reproduce ADH Proposition 99 with VanillaSC and CLUSTERSC

Source: https://mlsynth.readthedocs.io

Loads the California tobacco panel from the mlsynth repository, prepares the data with lagged-outcome predictors, and fits both a classical synthetic control (VanillaSC) and a robust OLS principal-component-regression control (CLUSTERSC). Prints the estimated average treatment effects for both methods.

```python
import pandas as pd
from mlsynth import VanillaSC, CLUSTERSC

# California tobacco panel, 1970-2000: per-capita cigarette sales plus the
# Abadie-Diamond-Hainmueller predictors.
url = ("https://raw.githubusercontent.com/jgreathouse9/mlsynth/"
       "main/basedata/augmented_cali_long.csv")
df = pd.read_csv(url)
df = df[df["year"] <= 2000].copy()
df["treated"] = ((df["state"] == "California")
                 & (df["year"] >= 1989)).astype(int)

# Three lagged-outcome predictors, as in ADH (2010), Table 2.
for L in (1975, 1980, 1988):
    df[f"cig{L}"] = df["state"].map(
        df[df["year"] == L].set_index("state")["cigsale"])

covariates = ["loginc", "p_cig", "pct15-24", "pc_beer",
              "cig1975", "cig1980", "cig1988"]
windows = {"loginc": (1980, 1988), "p_cig": (1980, 1988),
           "pct15-24": (1980, 1988), "pc_beer": (1984, 1988),
           "cig1975": (1975, 1975), "cig1980": (1980, 1980),
           "cig1988": (1988, 1988)}

common = dict(df=df, outcome="cigsale", treat="treated",
              unitid="state", time="year", display_graphs=False)

# Classical synthetic control on the canonical predictor set.
sc = VanillaSC({**common, "covariates": covariates,
                "covariate_windows": windows,
                "backend": "mscmt", "canonical_v": "min.loss.w",
                "seed": 0}).fit()

# Robust principal-component-regression control on the same outcomes.
pcr = CLUSTERSC({**common, "method": "pcr",
                 "pcr_objective": "OLS"}).fit()

print(f"VanillaSC ATT = {sc.att:+.2f} packs/yr")
print(f"OLS-PCR   ATT = {pcr.att:+.2f} packs/yr")
```

```text
VanillaSC ATT = -18.98 packs/yr
OLS-PCR   ATT = -21.39 packs/yr
```
Tokens:
Raw