Analyze a real public dataset

df = pd.read_csv(DATA_DIR / "public" / "penguins.csv", na_values=["NA"])
result = {
    "total_rows": len(df),
    "rows_with_any_missing": int(df.isna().any(axis=1).sum()),
    "complete_rows": len(df.dropna()),
    "rows_with_known_mass": int(df["body_mass_g"].notna().sum()),
}
print(result)
print("The original DataFrame still has", len(df), "rows.")

There are 344 rows, 11 with at least one missing field, 333 complete rows, and 342 with known body mass. Complete-case filtering would discard nine additional known body masses because another field is missing.

df = pd.read_csv(DATA_DIR / "public" / "penguins.csv", na_values=["NA"])
df.insert(0, "source_row", np.arange(1, len(df) + 1))
required = ["species", "sex", "body_mass_g"]
missing_required = df[required].isna().any(axis=1)
audit = df.loc[missing_required, ["source_row"] + required].copy()
result = df.loc[~missing_required].copy()
print("Rows retained:", len(result))
print("Rows excluded:", len(audit))
print(audit)

The revised question retains 333 records and excludes 11. The decision changed because the requested grouping now needs sex; it is not a rule that every analysis must use complete cases.

df = pd.read_csv(DATA_DIR / "public" / "penguins.csv", na_values=["NA"])
result = df.groupby("island", as_index=False).agg(Observations=("island", "size"), Known_mass=("body_mass_g", "count"), Mean_mass_g=("body_mass_g", "mean"))
result["Missing_mass"] = result["Observations"] - result["Known_mass"]
assert result["Observations"].sum() == len(df)
assert result["Known_mass"].sum() == df["body_mass_g"].count()
print(result.round({"Mean_mass_g": 2}))

Biscoe has 168 observations and 167 known masses; Dream has 124 and 124; Torgersen has 52 and 51. Their mean masses are approximately 4,716.02 g, 3,712.90 g, and 3,706.37 g. Different species composition is one reason to avoid interpreting these as an isolated island effect.

df = pd.read_csv(DATA_DIR / "public" / "penguins.csv", na_values=["NA"])
result = df.groupby("species", as_index=False).agg(Observations=("species", "size"), Known_length=("flipper_length_mm", "count"), Median_flipper_mm=("flipper_length_mm", "median"))
result["Missing_length"] = result["Observations"] - result["Known_length"]
fig, ax = plt.subplots(figsize=(7, 4))
ax.bar(result["species"], result["Median_flipper_mm"], color="#326f79")
for i, row in result.iterrows():
    ax.text(i, row["Median_flipper_mm"] + 3, f"n={int(row['Known_length'])}", ha="center")
ax.set(title="Median observed penguin flipper length, 2007–2009", xlabel="Species", ylabel="Median flipper length (mm)", ylim=(0, 250))
fig.tight_layout()
csv_path = OUTPUT_DIR / "public_penguins_flipper_summary.csv"
plot_path = OUTPUT_DIR / "public_penguins_flipper_summary.png"
result.to_csv(csv_path, index=False)
fig.savefig(plot_path, dpi=160)
print(result)
print("Source: Palmer Penguins; DOI 10.5281/zenodo.3960218. Medians use known lengths.")
print("The chart describes the available study observations; it is not a causal analysis.")
plt.show()

The median flipper lengths are 190 mm for Adelie, 196 mm for Chinstrap, and 216 mm for Gentoo, based on 151, 68, and 123 known measurements. The table and chart use the same units and denominators.