Link to notebook: balance_quickstart_rake.ipynb

Raking from known marginal totals

If you already have target marginal totals rather than row-level target data, rake() can build the synthetic target input for you via target_margins. Each variable's dictionary must sum to the same positive target total. For example:

import pandas as pd
from balance.weighting_methods.rake import rake

sample_df = pd.DataFrame({
    "gender": ["F", "M", "M", "F"],
    "region": ["N", "N", "S", "S"],
})
sample_weights = pd.Series([1.0, 1.0, 1.0, 1.0])

adjusted = rake(
    sample_df,
    sample_weights,
    target_df=None,
    target_weights=None,
    target_margins={
        "gender": {"F": 60.0, "M": 40.0},
        "region": {"N": 50.0, "S": 50.0},
    },
)

# The fitted weights are rescaled to the common target total (60 + 40 = 100).
round(float(adjusted["weight"].sum()), 6)
# 100.0

Use a larger target_margins_max_length when very small positive categories need more synthetic rows to be represented.