A Coding Guide to Survey Bias Correction Using Facebook Research Balance with IPW CBPS Ranking and Post Stratification Methods

import subprocess, sys
subprocess.check_call([sys.executable, "-m", "pip", "install", "-q", "balance"])


import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import warnings
warnings.filterwarnings("ignore")


from stability import Sample


np.random.seed(2024)
sns.set_theme(type="whitegrid", context="pocket book")

def simulate_population(n=50_000):
   age = np.clip(np.random.regular(45, 17, n), 18, 90).astype(int)
   gender = np.random.alternative(["M", "F"], dimension=n, p=[0.49, 0.51])
   schooling = np.random.alternative(
       ["HS", "SomeCollege", "Bachelor", "Graduate"],
       dimension=n, p=[0.35, 0.25, 0.25, 0.15],
   )
   revenue = np.exp(np.random.regular(10.5, 0.5, n))
   area = np.random.alternative(
       ["Urban", "Suburban", "Rural"], dimension=n, p=[0.40, 0.35, 0.25]
   )
   happiness = (
       50
       + 0.20 * (age - 45)
       + (schooling == "Graduate") * 8
       + (schooling == "Bachelor") * 4
       + (area == "Urban") * 3
       + np.log(revenue) * 2
       + np.random.regular(0, 5, n)
   )
   return pd.DataBody({
       "id": np.arange(n).astype(str),
       "age": age,
       "gender": gender,
       "schooling": schooling,
       "revenue": revenue.spherical(2),
       "area": area,
       "happiness": happiness.spherical(2),
   })


def biased_sample(pop, n=2_000):
   rating = (
       -0.04 * (pop["age"] - 30)
       + (pop["education"] == "Graduate") * 1.0
       + (pop["education"] == "Bachelor") * 0.6
       + (pop["region"] == "Urban") * 0.7
       - (pop["region"] == "Rural") * 0.5
   )
   p = 1 / (1 + np.exp(-score))
   p = p / p.sum()
   idx = np.random.alternative(pop.index, dimension=n, substitute=False, p=p)
   return pop.loc[idx].reset_index(drop=True)


target_df = simulate_population(50_000)
sample_df = biased_sample(target_df, 2_000)
target_for_balance = target_df.drop(columns=["happiness"])


print(f"Sample dimension : {len(sample_df):,}")
print(f"Target dimension : {len(target_for_balance):,}")
print(f"nTRUE inhabitants imply happiness : {target_df['happiness'].imply():.2f}")
print(f"Naive pattern imply happiness    : {sample_df['happiness'].imply():.2f}   <-- biased!")

pattern = Sample.from_frame(
   sample_df, id_column="id", outcome_columns=["happiness"]
)
goal = Sample.from_frame(target_for_balance, id_column="id")
sample_with_target = pattern.set_target(goal)


print("n--- Sample object ---")
print(sample_with_target)


print("n" + "=" * 60)
print(" PRE-ADJUSTMENT DIAGNOSTICS")
print("=" * 60)


asmd_before = sample_with_target.covars().asmd()
print("nASMD (Absolute Standardized Mean Difference) — decrease = higher stability")
print("Rule of thumb: |ASMD| > 0.10 signifies significant imbalance.")
print(asmd_before.T.spherical(3))


print("nMean of covariates (pattern vs goal):")
print(sample_with_target.covars().imply().T.spherical(3))

print("n" + "=" * 60)
print(" FITTING WEIGHTS — 4 METHODS")
print("=" * 60)


print("n>>> [1/4] IPW with LASSO logistic regression")
adjusted_ipw = sample_with_target.regulate(methodology="ipw")
print(adjusted_ipw.abstract())


print("n>>> [2/4] CBPS — Covariate Balancing Propensity Score")
strive:
   adjusted_cbps = sample_with_target.regulate(methodology="cbps")
   print(adjusted_cbps.abstract())
besides Exception as e:
   print("CBPS failed (skipping):", e)
   adjusted_cbps = None


print("n>>> [3/4] Raking (iterative proportional becoming)")
adjusted_rake = sample_with_target.regulate(methodology="rake")
print(adjusted_rake.abstract())


print("n>>> [4/4] Post-stratification (categoricals solely)")
cat_cols = ["id", "gender", "education", "region"]
sample_cat = Sample.from_frame(
   sample_df[cat_cols + ["happiness"]],
   id_column="id", outcome_columns=["happiness"],
)
target_cat = Sample.from_frame(target_for_balance[cat_cols], id_column="id")
adjusted_post = sample_cat.set_target(target_cat).regulate(methodology="poststratify")
print(adjusted_post.abstract())


print("n" + "=" * 60)
print(" METHOD COMPARISON")
print("=" * 60)


strategies = {
   "IPW":       adjusted_ipw,
   "CBPS":      adjusted_cbps,
   "Rake":      adjusted_rake,
   "PostStrat": adjusted_post,
}


def safe_mean_asmd(asmd_df, desire="self"):
   """Mean ASMD throughout covariates from a stability asmd DataBody."""
   row = desire if desire in asmd_df.index else asmd_df.index[0]
   if "imply(asmd)" in asmd_df.columns:
       return float(asmd_df.loc[row, "mean(asmd)"])
   return float(asmd_df.loc[row].imply())


asmd_means   = {"Unadjusted": safe_mean_asmd(asmd_before)}
outcome_means = {"Naive pattern": float(sample_df["happiness"].imply())}
deff_vals    = {}


for identify, m in strategies.gadgets():
   if m is None:
       proceed
   asmd_means[name]   = safe_mean_asmd(m.covars().asmd(), desire="self")
   outcome_means[name] = float(m.outcomes().imply()["happiness"].iloc[0])
   w = m.to_df()["weight"].values
   deff_vals[name] = (w.sum() ** 2) / (len(w) * np.sum(w ** 2))
outcome_means["TRUE pop"] = float(target_df["happiness"].imply())


print("nMean ASMD throughout covariates (decrease = higher stability):")
for okay, v in asmd_means.gadgets():
   print(f"  {okay:14s}: {v:.4f}")


print("nWeighted estimate of imply happiness:")
for okay, v in outcome_means.gadgets():
   print(f"  {okay:14s}: {v:.3f}")


print("nKish's efficient sample-size ratio (1.0 = no data loss):")
for okay, v in deff_vals.gadgets():
   print(f"  {okay:14s}: {v:.3f}   (n_eff ≈ {int(v * len(sample_df))})")

fig, axes = plt.subplots(2, 2, figsize=(14, 10))


colors_a = ["gray", "#1f77b4", "#ff7f0e", "#2ca02c", "#d62728"][: len(asmd_means)]
axes[0, 0].bar(record(asmd_means.keys()), record(asmd_means.values()), coloration=colors_a)
axes[0, 0].axhline(0.1, ls="--", coloration="purple", label="0.10 imbalance threshold")
axes[0, 0].set_title("Mean ASMD throughout covariates")
axes[0, 0].set_ylabel("Mean ASMD"); axes[0, 0].legend()
axes[0, 0].tick_params(axis="x", rotation=20)


fact = target_df["happiness"].imply()
colors_b = ["#888"] + ["#1f77b4", "#ff7f0e", "#2ca02c", "#d62728"][: len(methods)] + ["black"]
axes[0, 1].bar(record(outcome_means.keys()), record(outcome_means.values()),
              coloration=colors_b[: len(outcome_means)])
axes[0, 1].axhline(fact, ls="--", coloration="black", label=f"fact = {fact:.2f}")
axes[0, 1].set_title("Estimated imply happiness vs floor fact")
axes[0, 1].set_ylabel("Mean happiness"); axes[0, 1].legend()
axes[0, 1].tick_params(axis="x", rotation=20)


w_ipw = adjusted_ipw.to_df()["weight"].values
axes[1, 0].hist(w_ipw, bins=40, coloration="steelblue", edgecolor="white")
axes[1, 0].set_title(
   f"IPW weight distributionn"
   f"min={w_ipw.min():.2f}  median={np.median(w_ipw):.2f}  max={w_ipw.max():.2f}"
)
axes[1, 0].set_xlabel("weight"); axes[1, 0].set_ylabel("depend")


ages = sample_df["age"].values
bins = np.linspace(18, 90, 31)
axes[1, 1].hist(target_df["age"], bins=bins, density=True, alpha=0.45,
               coloration="inexperienced", label="Target (fact)")
axes[1, 1].hist(ages, bins=bins, density=True, alpha=0.45,
               coloration="purple", label="Sample (biased)")
axes[1, 1].hist(ages, bins=bins, density=True, alpha=0.45,
               coloration="blue", weights=w_ipw, label="Sample (IPW-weighted)")
axes[1, 1].set_title("Age distribution: bias correction by IPW")
axes[1, 1].set_xlabel("Age"); axes[1, 1].set_ylabel("density"); axes[1, 1].legend()


plt.tight_layout()
plt.savefig("balance_diagnostics.png", dpi=110, bbox_inches="tight")
plt.present()


print("n" + "=" * 60)
print(" ADVANCED — controlling variance with max_de")
print("=" * 60)
print("max_de=1.5 trims excessive weights so the design impact stays ≤ 1.5,")
print("buying and selling slightly bias for tighter confidence intervals.n")
adjusted_trim = sample_with_target.regulate(methodology="ipw", max_de=1.5)
print(adjusted_trim.abstract())


out = adjusted_ipw.to_df()
out.to_csv("balance_weighted_sample.csv", index=False)
print("nSaved weighted pattern → balance_weighted_sample.csv")
print("Saved diagnostics plot → balance_diagnostics.png")
print("nFirst 5 rows of weighted output:")
print(out.head())


err_naive = abs(sample_df["happiness"].imply() - fact)
err_ipw   = abs(outcome_means["IPW"]            - fact)
print("n" + "=" * 60)
print(" BIAS REDUCTION SUMMARY")
print("=" * 60)
print(f"Naive estimator error : {err_naive:.3f}")
print(f"IPW estimator error   : {err_ipw:.3f}")
print(f"Bias discount        : {(1 - err_ipw / max(err_naive, 1e-9)) * 100:.1f}%")