import hashlib, re, warnings
from pathlib import Path

import numpy as np
import pandas as pd
import matplotlib as mpl
import matplotlib.pyplot as plt
from scipy.stats import chisquare

warnings.filterwarnings("ignore", category=FutureWarning)
pd.set_option("display.width", 200)
pd.set_option("display.max_columns", 60)

HERE     = Path.cwd()
RAW      = HERE.parent / "gaza data"      # the MoH lists (not in the repo)
POP_CSV  = HERE / "data" / "pse_admpop_adm1_2023.csv"
OUT_JS   = HERE.parent / "assets" / "gaza-data.js"

# --- the site's palette, so the notebook and the published charts agree -------
# Navy ink on cream, marigold accent. SAND is the reference series (the
# living population) and MARK the one highlighted series (the dead); the
# pair was checked for separation rather than picked by eye.
PAPER  = "#FFFBF0"   # the chart surface, matching the site's chart cards
SAND   = "#C0AE86"   # the population baseline: context, not a category
SAND_L = "#D3C4A2"
MARK   = "#1E4A85"   # the dead: the one highlighted series
BRICK  = "#BE3C2C"   # the single hot accent, used for the ratio line
MARI   = "#F2C230"
INK    = "#12305C"
MUTED  = "#5C6779"

mpl.rcParams.update({
    "figure.dpi": 130, "savefig.dpi": 130,
    "font.family": "sans-serif", "font.style": "normal",
    # Deliberately NOT "Roboto", even though the site's pages use it: whether
    # a machine has a usable upright Roboto installed varies, and matplotlib
    # will happily match an italic-only install and render the whole figure
    # in italics. A neutral grotesque renders the same everywhere.
    "font.sans-serif": ["Helvetica Neue", "Helvetica", "Arial", "DejaVu Sans"],
    "font.size": 9,
    "axes.edgecolor": "#E5DAC0", "axes.labelcolor": MUTED, "axes.titlesize": 11,
    "axes.titleweight": "bold", "axes.titlecolor": INK, "axes.titlelocation": "left",
    "axes.spines.top": False, "axes.spines.right": False,
    "xtick.color": MUTED, "ytick.color": MUTED,
    "xtick.labelcolor": MUTED, "ytick.labelcolor": MUTED,
    "text.color": INK,
    "figure.facecolor": PAPER, "axes.facecolor": PAPER, "savefig.facecolor": PAPER,
    "legend.frameon": False, "legend.fontsize": 8.5,
})

# Paths are printed relative to the repository root: this notebook is
# published as HTML, and nobody needs my home directory.
rel = lambda p: p.relative_to(HERE.parent)
print(f"raw lists folder : {rel(RAW)}/  -- {'found' if RAW.is_dir() else 'MISSING'}")
print(f"population file  : {rel(POP_CSV)} -- {'found' if POP_CSV.is_file() else 'MISSING'}")

raw lists folder : gaza data/  -- found
population file  : analysis/data/pse_admpop_adm1_2023.csv -- found

files = sorted(RAW.glob("*.csv"))
rows, seen, lists = [], {}, []

for f in files:
    digest = hashlib.md5(f.read_bytes()).hexdigest()
    date   = re.match(r"MoH-(\d{4}-\d{2}-\d{2})", f.name).group(1)
    dup_of = seen.get(digest)
    rows.append({"file": f.name[:34] + "…", "list date": date,
                 "md5": digest[:8], "duplicate of": dup_of or ""})
    if dup_of:
        continue
    seen[digest] = date
    df = pd.read_csv(f, dtype=str, keep_default_na=False)
    lists.append((date, df))

inventory = pd.DataFrame(rows)
display(inventory)
print(f"\n{len(files)} files → {len(lists)} distinct lists "
      f"({len(files) - len(lists)} byte-identical duplicate).")

10 files → 9 distinct lists (1 byte-identical duplicate).

for date, df in lists:
    print(f"{date}  n={len(df):>6,}  columns: {', '.join(df.columns)}")

a = dict(lists)["2025-03-23"]
b = dict(lists)["2025-07-31"]
m = a[["id", "age", "dob", "sex"]].merge(b[["id", "age", "dob", "sex"]],
                                         on="id", suffixes=("_mar", "_jul"))
print(f"\n{len(m):,} people appear in both the 2025-03-23 and 2025-07-31 lists.")
for col in ("age", "dob", "sex"):
    same = (m[f"{col}_mar"].str.lower() == m[f"{col}_jul"].str.lower()).mean()
    print(f"  {col:4s} unchanged between the two lists: {same*100:6.2f}%")

2024-01-05  n=14,140  columns: id, name_ar_raw, dob, sex
2024-03-29  n=20,390  columns: id, name_ar_raw, dob, age, sex, source
2024-04-30  n=24,669  columns: id, name_ar_raw, dob, age, sex, source
2024-06-30  n=28,173  columns: index, name_ar_raw, id, dob, sex, age, source
2024-08-31  n=34,306  columns: index, name_ar_raw, id, dob, sex, age, source
2025-03-23  n=50,020  columns: index, name_en, name_ar_raw, age, dob, sex, id, source
2025-06-15  n=55,202  columns: index, name_en, name_ar_raw, age, dob, sex, id, source
2025-07-15  n=58,380  columns: index, name_en, name_ar_raw, age, dob, sex, id, source
2025-07-31  n=60,199  columns: index, name_en, name_ar_raw, age, dob, sex, id, source

49,947 people appear in both the 2025-03-23 and 2025-07-31 lists.
  age  unchanged between the two lists:  99.72%
  dob  unchanged between the two lists:  99.99%
  sex  unchanged between the two lists: 100.00%

latest_date, latest = lists[-1]
implied = pd.to_datetime(latest["dob"]) + pd.to_timedelta(
    pd.to_numeric(latest["age"]) * 365.2425, unit="D")
print(f"list published {latest_date}; dob + age implies a date of:")
print(implied.quantile([.05, .25, .5, .75, .95]).dt.date.to_string())

list published 2025-07-31; dob + age implies a date of:
0.05    2022-12-23
0.25    2023-05-07
0.50    2023-09-28
0.75    2024-04-15
0.95    2025-01-26

BANDS = ["0-4", "5-9", "10-14", "15-19", "20-24", "25-29", "30-34",
         "35-39", "40-44", "45-49", "50-54", "55-59", "60-64", "65+"]
EDGES = [0, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 200]


def clean(date, df):
    # -> (usable rows with age + sex, number excluded, how age was obtained)
    sex = df["sex"].str.strip().str.lower()
    if "age" in df.columns:
        age, how = pd.to_numeric(df["age"], errors="coerce"), "age column"
    else:
        dob = pd.to_datetime(df["dob"], errors="coerce")
        age = np.floor((pd.Timestamp(date) - dob).dt.days / 365.2425)
        how = "derived from dob"
    ok = age.notna() & age.between(0, 120) & sex.isin(["m", "f"])
    out = pd.DataFrame({"id": df["id"], "age": age[ok], "sex": sex[ok]}).dropna(subset=["age"])
    out["band"] = pd.cut(out["age"], bins=EDGES, right=False, labels=BANDS)
    return out, int((~ok).sum()), how


quality = []
clean_lists = []
for date, df in lists:
    c, excluded, how = clean(date, df)
    clean_lists.append((date, c))
    quality.append({"list": date, "records": len(df), "usable": len(c),
                    "excluded": excluded, "age from": how,
                    "dob = 1990-01-01": int((df["dob"] == "1990-01-01").sum())})

display(pd.DataFrame(quality).set_index("list"))

churn = []
for (d0, a), (d1, b) in zip(lists, lists[1:]):
    ids_a, ids_b = set(a["id"]), set(b["id"])
    churn.append({"from": d0, "to": d1,
                  "added": len(ids_b - ids_a), "removed": len(ids_a - ids_b),
                  "net": len(ids_b) - len(ids_a)})
display(pd.DataFrame(churn))

pop_raw = pd.read_csv(POP_CSV)
gaza = pop_raw[pop_raw["ADM1_PCODE"] == "PS02"].iloc[0]

SRC17 = ["00_04", "05_09", "10_14", "15_19", "20_24", "25_29", "30_34", "35_39", "40_44",
         "45_49", "50_54", "55_59", "60_64", "65_69", "70_74", "75_79", "80Plus"]

def collapse(prefix):
    v = [int(gaza[f"{prefix}_{b}"]) for b in SRC17]
    return v[:13] + [sum(v[13:])]          # 65-69, 70-74, 75-79, 80+  ->  65+

pop_m, pop_f = collapse("M"), collapse("F")

population = pd.DataFrame({"male": pop_m, "female": pop_f}, index=BANDS)
population["total"] = population.sum(axis=1)
display(population)

assert population["total"].sum() == int(gaza["T_TL"]), "collapse lost people"
print(f"\nGaza Strip, {int(gaza['year'])}: {population['total'].sum():,} people "
      f"({sum(pop_m):,} male, {sum(pop_f):,} female)")
print(f"population sex ratio: {sum(pop_m)/sum(pop_f):.3f} males per female")
print(f"under 15: {population['total'][:3].sum()/population['total'].sum()*100:.1f}% of the population")

Gaza Strip, 2023: 2,226,544 people (1,127,926 male, 1,098,618 female)
population sex ratio: 1.027 males per female
under 15: 40.4% of the population

LIST_DATE, latest_clean = clean_lists[-1]
tab = latest_clean.pivot_table(index="band", columns="sex", aggfunc="size",
                               fill_value=0, observed=False)
deaths_m = [int(tab.loc[b, "m"]) for b in BANDS]
deaths_f = [int(tab.loc[b, "f"]) for b in BANDS]
D, P = sum(deaths_m) + sum(deaths_f), sum(pop_m) + sum(pop_f)

comp = pd.DataFrame({
    "pop":          population["total"],
    "pop %":        population["total"] / P * 100,
    "deaths":       np.array(deaths_m) + np.array(deaths_f),
    "deaths %":     (np.array(deaths_m) + np.array(deaths_f)) / D * 100,
}, index=BANDS)
comp["expected if random"] = comp["pop %"] / 100 * D
comp["actual ÷ expected"]  = comp["deaths"] / comp["expected if random"]
display(comp.round(2))

print(f"list: Gaza MoH named dead, {LIST_DATE} — {D:,} usable records")

list: Gaza MoH named dead, 2025-07-31 — 60,199 usable records

def group(lo, hi, sex=None, label=""):
    idx   = range(lo, hi + 1)
    sexes = ["m", "f"] if sex is None else [sex]
    src   = {"m": (deaths_m, pop_m), "f": (deaths_f, pop_f)}
    act   = sum(src[s][0][i] for s in sexes for i in idx)
    pop   = sum(src[s][1][i] for s in sexes for i in idx)
    exp   = pop / P * D
    return {"group": label, "pop": pop, "pop %": pop / P * 100,
            "actual deaths": act, "deaths %": act / D * 100,
            "expected if random": exp, "actual ÷ expected": act / exp}

headline = pd.DataFrame([
    group(0,  2,  None, "Children under 15"),
    group(3, 11,  "m",  "Males 15-59"),
    group(3, 11,  "f",  "Females 15-59"),
    group(5, 10,  "m",  "Males 25-54"),
    group(5, 10,  "f",  "Females 25-54"),
    group(13, 13, None, "Aged 65 and over"),
]).set_index("group")
display(headline.round(2))

obs = np.array(deaths_m + deaths_f, dtype=float)
exp = np.array(pop_m + pop_f, dtype=float)
exp = exp / exp.sum() * obs.sum()
stat, p = chisquare(obs, exp)
print(f"chi-square = {stat:,.0f}   df = {len(obs)-1}   p = {p:.3g}")
print("(A p of 0 here means 'below floating-point resolution', not 'exactly zero'.)\n")

labels  = [f"male {b}" for b in BANDS] + [f"female {b}" for b in BANDS]
contrib = (obs - exp) ** 2 / exp
print("cells contributing most of the statistic:")
for i in np.argsort(-contrib)[:6]:
    print(f"  {labels[i]:14s} observed {obs[i]:7,.0f}   expected {exp[i]:7,.0f}   "
          f"ratio {obs[i]/exp[i]:.2f}")

chi-square = 19,308   df = 27   p = 0
(A p of 0 here means 'below floating-point resolution', not 'exactly zero'.)

cells contributing most of the statistic:
  male 30-34     observed   5,109   expected   2,367   ratio 2.16
  male 35-39     observed   3,833   expected   1,735   ratio 2.21
  male 25-29     observed   4,984   expected   2,628   ratio 1.90
  male 20-24     observed   4,730   expected   2,652   ratio 1.78
  female 0-4     observed   2,241   expected   4,474   ratio 0.50
  male 65+       observed   1,791   expected     834   ratio 2.15

share = lambda v, tot: np.array(v) / tot * 100
dm, df_, pm, pf = (share(deaths_m, D), share(deaths_f, D),
                   share(pop_m, P),    share(pop_f, P))

fig, (axl, axr) = plt.subplots(1, 2, figsize=(9, 5), sharey=True,
                               gridspec_kw={"wspace": .30})
y   = np.arange(len(BANDS))
lim = max(dm.max(), df_.max(), pm.max(), pf.max()) * 1.15

for ax, pop_s, dth_s, title in ((axl, pm, dm, "MALE"), (axr, pf, df_, "FEMALE")):
    ax.barh(y, pop_s, height=.80, color=SAND_L, label="share of population", zorder=2)
    ax.barh(y, dth_s, height=.44, color=MARK,   label="share of the dead",   zorder=3)
    ax.set_title(title, fontsize=9, color=MUTED, pad=8)
    ax.set_xlim(0, lim)
    ax.set_xlabel("% of that sex-and-age group's own total")
    ax.grid(axis="x", color="#EFE5CD", lw=.8, zorder=0)
    ax.set_axisbelow(True)

axl.invert_xaxis()
axl.set_yticks(y, BANDS)
axl.tick_params(axis="y", length=0)
axr.tick_params(axis="y", length=0, labelleft=False)
fig.suptitle("The dead are not a cross-section of the living",
             x=.02, y=1.10, ha="left", fontsize=12, fontweight="bold", color=INK)
fig.legend(*axl.get_legend_handles_labels(), loc="upper left",
           bbox_to_anchor=(.02, 1.045), ncols=2)
fig.text(.02, -.02, f"Gaza MoH named list, {LIST_DATE} (n = {D:,}) · "
                    "population: PCBS 2023 via OCHA/UNFPA COD",
         fontsize=7.5, color=MUTED)
plt.show()

groups = [group(0, 2, None, "Under 15"),
          group(3, 11, "m", "Males 15-59"),
          group(5, 10, "m", "Males 25-54"),
          group(5, 10, "f", "Females 25-54")]

fig, ax = plt.subplots(figsize=(8, 3.6))
y = np.arange(len(groups))
ax.barh(y + .19, [g["expected if random"] for g in groups], height=.34,
        color=SAND, label="expected if deaths were drawn at random")
ax.barh(y - .19, [g["actual deaths"] for g in groups], height=.34,
        color=MARK, label="actual")

for i, g in enumerate(groups):                       # direct labels: the grey needs them
    ax.text(g["expected if random"] + D * .006, i + .19, f'{g["expected if random"]:,.0f}',
            va="center", fontsize=8.5, color=MUTED)
    ax.text(g["actual deaths"] + D * .006, i - .19, f'{g["actual deaths"]:,.0f}',
            va="center", fontsize=8.5, color=INK, fontweight="bold")

ax.set_yticks(y, [f'{g["group"]}\n{g["pop %"]:.1f}% of the population' for g in groups])
ax.tick_params(axis="y", length=0)
ax.invert_yaxis()
ax.set_xlim(0, max(max(g["expected if random"], g["actual deaths"]) for g in groups) * 1.22)
ax.set_xlabel("named deaths")
ax.grid(axis="x", color="#EFE5CD", lw=.8)
ax.set_axisbelow(True)
ax.legend(loc="lower right")
ax.set_title("Expected against actual", pad=10)
plt.show()

for g in groups:
    print(f'{g["group"]:16s} {g["actual deaths"]:6,.0f} actual vs {g["expected if random"]:7,.0f} '
          f'expected  =  {g["actual ÷ expected"]:.2f}×')

Under 15         14,824 actual vs  24,298 expected  =  0.61×
Males 15-59      30,082 actual vs  16,708 expected  =  1.80×
Males 25-54      19,702 actual vs  10,020 expected  =  1.97×
Females 25-54     6,815 actual vs   9,924 expected  =  0.69×

ratio     = np.array([m_ / f_ if f_ else np.nan for m_, f_ in zip(deaths_m, deaths_f)])
pop_ratio = np.array(pop_m) / np.array(pop_f)

sex_tbl = pd.DataFrame({"male deaths": deaths_m, "female deaths": deaths_f,
                        "M:F deaths": ratio.round(2),
                        "M:F population": pop_ratio.round(3)}, index=BANDS)
display(sex_tbl)

fig, ax = plt.subplots(figsize=(8, 4))
x = np.arange(len(BANDS))
ax.plot(x, pop_ratio, color=SAND, lw=2, ls="--", zorder=2,
        label="M:F in the living population")
ax.plot(x, ratio, color=BRICK, lw=2, marker="o", ms=5, zorder=3,
        label="M:F among the named dead")
ax.axhspan(0, 1.10, color="#F5ECD6", zorder=0)

peak = int(np.nanargmax(ratio))
ax.annotate(f"{ratio[peak]:.2f}", (peak, ratio[peak]), textcoords="offset points",
            xytext=(0, 9), ha="center", fontsize=8.5, fontweight="bold", color=INK)
ax.annotate(f"{ratio[0]:.2f}", (0, ratio[0]), textcoords="offset points",
            xytext=(4, -14), fontsize=8.5, color=INK)

ax.set_xticks(x, BANDS, rotation=45, ha="right")
ax.set_ylim(0, max(np.nanmax(ratio) * 1.18, 3.6))
ax.set_ylabel("male deaths per female death")
ax.set_xlabel("age band")
ax.grid(axis="y", color="#EFE5CD", lw=.8)
ax.set_axisbelow(True)
ax.legend(loc="upper right")
ax.set_title("The ratio climbs with age, then falls back", pad=10)
plt.show()

print(f"0-4:    {ratio[0]:.2f} male deaths per female death "
      f"(population: {pop_ratio[0]:.2f})")
print(f"peak:   {ratio[peak]:.2f} at {BANDS[peak]}")
print(f"65+:    {ratio[-1]:.2f}")
print(f"whole list: {sum(deaths_m)/sum(deaths_f):.2f}")

0-4:    1.12 male deaths per female death (population: 1.04)
peak:   3.28 at 35-39
65+:    1.58
whole list: 2.13

trend = []
for date, c in clean_lists:
    trend.append({
        "list": date, "usable": len(c),
        "% female":        (c.sex == "f").mean() * 100,
        "% under 15":      (c.age < 15).mean() * 100,
        "% males 15-59":   ((c.sex == "m") & c.age.between(15, 59)).mean() * 100,
        "M:F ratio":       (c.sex == "m").sum() / (c.sex == "f").sum(),
    })
trend = pd.DataFrame(trend).set_index("list")
display(trend.round(2))

fig, (a1, a2) = plt.subplots(1, 2, figsize=(9.5, 3.8), gridspec_kw={"wspace": .28})
x = np.arange(len(trend))

a1.plot(x, trend["% males 15-59"], color=MARK, lw=2, marker="o", ms=5, label="males 15–59")
a1.plot(x, trend["% under 15"],    color=SAND, lw=2, marker="o", ms=5, label="children under 15")
a1.axhline(27.8, color=MARK, lw=1, ls=":", zorder=1)
a1.axhline(40.4, color=SAND, lw=1, ls=":", zorder=1)
a1.text(len(trend) - 1, 29.4, "27.8% of the population", ha="right", fontsize=7.5, color=MARK)
a1.text(len(trend) - 1, 41.9, "40.4% of the population", ha="right", fontsize=7.5, color=MUTED)
a1.set_ylim(0, 58)
a1.set_ylabel("% of the named dead")
a1.legend(loc="lower left")
a1.set_title("Composition of each list", pad=10)

a2.plot(x, trend["M:F ratio"], color=BRICK, lw=2, marker="o", ms=5)
a2.axhline(1.03, color=SAND, lw=1, ls="--")
a2.text(0, 1.13, "population ratio 1.03", fontsize=7.5, color=MUTED)
a2.set_ylim(0, 2.5)
a2.set_ylabel("male deaths per female death")
a2.set_title("Sex ratio of each list", pad=10)

for ax in (a1, a2):
    ax.set_xticks(x, trend.index, rotation=45, ha="right", fontsize=7.5)
    ax.grid(axis="y", color="#EFE5CD", lw=.8)
    ax.set_axisbelow(True)

fig.text(.02, -.06, "First list (2024-01-05) has no age column; age derived from date of birth. "
                    "Lists are cumulative.", fontsize=7.5, color=MUTED)
plt.show()

batches = []
for (d0, a), (d1, b) in zip(clean_lists, clean_lists[1:]):
    new = b[~b["id"].isin(set(a["id"]))]
    if not len(new):
        continue
    batches.append({"names added in": f"{d0} → {d1}", "n": len(new),
                    "% female":      (new.sex == "f").mean() * 100,
                    "% under 15":    (new.age < 15).mean() * 100,
                    "% males 15-59": ((new.sex == "m") & new.age.between(15, 59)).mean() * 100,
                    "M:F ratio":     (new.sex == "m").sum() / (new.sex == "f").sum()})
batches = pd.DataFrame(batches).set_index("names added in")
display(batches.round(2))

print(f'every batch M:F between {batches["M:F ratio"].min():.2f} and '
      f'{batches["M:F ratio"].max():.2f}; population ratio is 1.03')

every batch M:F between 1.97 and 3.76; population ratio is 1.03

def js_arr(v):
    return "[" + ", ".join(str(x).rjust(6) for x in v) + "]"

HEAD = [
    "/* ------------------------------------------------------------------",
    "   gaza-data.js - counts behind the casualty-demography essay.",
    "",
    "   GENERATED FILE. Do not hand-edit.",
    "   Produced by analysis/gaza-casualty-demographics.ipynb; re-run that",
    "   notebook to regenerate. Every number below is a count, not an",
    "   estimate: `deaths` tallies named records in ONE published Gaza MoH",
    "   list, `pop` is a published PCBS population figure. The age bands of",
    "   the two match exactly, which is what makes the shares comparable.",
    "   ------------------------------------------------------------------ */",
    "",
    "const GAZA = {",
    "  loaded: true,",
    "",
    f"  source:    'Gaza MoH named list, {LIST_DATE}, n = {D:,}',",
    "  popSource: 'PCBS Gaza Strip, 2023, via OCHA/UNFPA COD cod-ps-pse',",
    f'  excluded:  {quality[-1]["excluded"]},    // records dropped for missing/unparseable age or sex',
    "",
    "  bands: ['" + "', '".join(BANDS[:7]) + "',",
    "          '" + "', '".join(BANDS[7:]) + "'],",
    "",
    "  deaths: {",
    f"    m: {js_arr(deaths_m)},",
    f"    f: {js_arr(deaths_f)}",
    "  },",
    "  pop: {",
    f"    m: {js_arr(pop_m)},",
    f"    f: {js_arr(pop_f)}",
    "  }",
    "};",
    "",
]
js = "\n".join(HEAD)

# keep the derived helpers already in the file; only the counts are regenerated
tail = OUT_JS.read_text().split("/* --- derived,")[1]
OUT_JS.write_text(js + "/* --- derived," + tail)

print(f"wrote {rel(OUT_JS)}  ({OUT_JS.stat().st_size:,} bytes)\n")
print(js[js.index("const GAZA"):])

wrote assets/gaza-data.js  (3,571 bytes)

const GAZA = {
  loaded: true,

  source:    'Gaza MoH named list, 2025-07-31, n = 60,199',
  popSource: 'PCBS Gaza Strip, 2023, via OCHA/UNFPA COD cod-ps-pse',
  excluded:  0,    // records dropped for missing/unparseable age or sex

  bands: ['0-4', '5-9', '10-14', '15-19', '20-24', '25-29', '30-34',
          '35-39', '40-44', '45-49', '50-54', '55-59', '60-64', '65+'],

  deaths: {
    m: [  2503,   2545,   3104,   4453,   4730,   4984,   5109,   3833,   2529,   1824,   1423,   1197,    965,   1791],
    f: [  2241,   2277,   2154,   1832,   1579,   1704,   1619,   1170,    890,    775,    657,    638,    536,   1137]
  },
  pop: {
    m: [171595, 145276, 141660, 120553,  98073,  97192,  87562,  64156,  49109,  41317,  31260,  28753,  20588,  30832],
    f: [165462, 139182, 135532, 115384,  93854,  94657,  85986,  64356,  49977,  41214,  30876,  27135,  19882,  35121]
  }
};

# Final self-check: the published file must reproduce the headline numbers.
txt = OUT_JS.read_text()
assert "loaded: true" in txt
assert f"n = {D:,}" in txt
print(f"total named dead in exported file : {sum(deaths_m) + sum(deaths_f):,}")
print(f"total population in exported file : {sum(pop_m) + sum(pop_f):,}")
print(f"children under 15, actual         : {sum(deaths_m[:3]) + sum(deaths_f[:3]):,}")
print(f"children under 15, if random      : {comp['expected if random'][:3].sum():,.0f}")
print(f"males 15-59, share of the dead    : {group(3,11,'m')['deaths %']:.1f}%")
print(f"males 15-59, share of Gaza        : {group(3,11,'m')['pop %']:.1f}%")

total named dead in exported file : 60,199
total population in exported file : 2,226,544
children under 15, actual         : 14,824
children under 15, if random      : 24,298
males 15-59, share of the dead    : 50.0%
males 15-59, share of Gaza        : 27.8%


Numerator	The Gaza Ministry of Health's own published lists of named dead, ten CSV files covering January 2024 to July 2025, archived at gazadeaths.org. Each row is one person: national ID, name, date of birth, age, sex.
Denominator	Gaza Strip population by five-year age band and sex, PCBS figures for reference year 2023, published as the OCHA/UNFPA Common Operational Dataset `cod-ps-pse`. This is a pre-war baseline, which is what the test needs.

	file	list date	md5	duplicate of
0	MoH-2024-01-05-rNQ2cecvsbOdzGN3EDf…	2024-01-05	ca7aba64
1	MoH-2024-03-29-tE8GTpcNl0sHnjN5Ygj…	2024-03-29	f751d170
2	MoH-2024-04-30_edited-FQ20GX6KVBCL…	2024-04-30	36a8b7c4
3	MoH-2024-04-30_edited-SFkYs2t6xSEs…	2024-04-30	36a8b7c4	2024-04-30
4	MoH-2024-06-30_edited-FeGee1fcc1Jn…	2024-06-30	1aaf9328
5	MoH-2024-08-31_edited-VJwKINol6A9q…	2024-08-31	5408ba2c
6	MoH-2025-03-23-NehFnLUa47bE42mkFsk…	2025-03-23	cc28a438
7	MoH-2025-06-15-K6mJgX3PJ1iZTWB8hJB…	2025-06-15	caeea485
8	MoH-2025-07-15-OR0Ge1rNshayOYuiQ9N…	2025-07-15	b6f6d4f6
9	MoH-2025-07-31-koX2irOztUhD4wldRZd…	2025-07-31	2d1cf608

	pop	pop %	deaths	deaths %	expected if random	actual ÷ expected
0-4	337057	15.14	4744	7.88	9113.00	0.52
5-9	284458	12.78	4822	8.01	7690.88	0.63
10-14	277192	12.45	5258	8.73	7494.43	0.70
15-19	235937	10.60	6285	10.44	6379.02	0.99
20-24	191927	8.62	6309	10.48	5189.12	1.22
25-29	191849	8.62	6688	11.11	5187.02	1.29
30-34	173548	7.79	6728	11.18	4692.21	1.43
35-39	128512	5.77	5003	8.31	3474.57	1.44
40-44	99086	4.45	3419	5.68	2678.99	1.28
45-49	82531	3.71	2599	4.32	2231.39	1.16
50-54	62136	2.79	2080	3.46	1679.97	1.24
55-59	55888	2.51	1835	3.05	1511.04	1.21
60-64	40470	1.82	1501	2.49	1094.19	1.37
65+	65953	2.96	2928	4.86	1783.17	1.64

	pop	pop %	actual deaths	deaths %	expected if random	actual ÷ expected
group
Children under 15	898707	40.36	14824	24.62	24298.31	0.61
Males 15-59	617975	27.75	30082	49.97	16708.17	1.80
Females 15-59	603439	27.10	10864	18.05	16315.16	0.67
Males 25-54	370596	16.64	19702	32.73	10019.79	1.97
Females 25-54	367066	16.49	6815	11.32	9924.35	0.69
Aged 65 and over	65953	2.96	2928	4.86	1783.17	1.64

	usable	% female	% under 15	% males 15-59	M:F ratio
list
2024-01-05	13143	43.19	31.66	36.54	1.32
2024-03-29	20042	39.73	26.31	42.28	1.52
2024-04-30	24240	37.36	24.88	44.72	1.68
2024-06-30	28173	36.96	27.32	43.14	1.71
2024-08-31	34306	36.00	26.81	44.10	1.78
2025-03-23	50020	32.99	25.32	48.46	2.03
2025-06-15	55202	32.73	25.14	48.91	2.06
2025-07-15	58380	32.08	24.71	49.76	2.12
2025-07-31	60199	31.91	24.62	49.97	2.13

Gaza Health Ministry named casualty lists — age and sex distribution¶

1. Inventory the lists¶

2. What the columns actually mean¶

3. Cleaning rule, stated once and applied everywhere¶

4. The denominator¶

5. The test¶

6. The distribution, drawn¶

7. Expected against actual¶

8. The sex ratio¶

9. Does the pattern hold across all nine lists?¶

10. What the notebook establishes, and what it does not¶

11. Export the counts to the site¶

	records	usable	excluded	age from	dob = 1990-01-01
list
2024-01-05	14140	13143	997	derived from dob	31
2024-03-29	20390	20042	348	age column	43
2024-04-30	24669	24240	429	age column	41
2024-06-30	28173	28173	0	age column	18
2024-08-31	34306	34306	0	age column	17
2025-03-23	50020	50020	0	age column	4
2025-06-15	55202	55202	0	age column	4
2025-07-15	58380	58380	0	age column	4
2025-07-31	60199	60199	0	age column	4

	from	to	added	removed	net
0	2024-01-05	2024-03-29	6445	195	6250
1	2024-03-29	2024-04-30	4463	184	4279
2	2024-04-30	2024-06-30	7431	3927	3504
3	2024-06-30	2024-08-31	6658	525	6133
4	2024-08-31	2025-03-23	18666	2952	15714
5	2025-03-23	2025-06-15	5244	62	5182
6	2025-06-15	2025-07-15	3195	17	3178
7	2025-07-15	2025-07-31	1826	7	1819

	male	female	total
0-4	171595	165462	337057
5-9	145276	139182	284458
10-14	141660	135532	277192
15-19	120553	115384	235937
20-24	98073	93854	191927
25-29	97192	94657	191849
30-34	87562	85986	173548
35-39	64156	64356	128512
40-44	49109	49977	99086
45-49	41317	41214	82531
50-54	31260	30876	62136
55-59	28753	27135	55888
60-64	20588	19882	40470
65+	30832	35121	65953

	male deaths	female deaths	M:F deaths	M:F population
0-4	2503	2241	1.12	1.037
5-9	2545	2277	1.12	1.044
10-14	3104	2154	1.44	1.045
15-19	4453	1832	2.43	1.045
20-24	4730	1579	3.00	1.045
25-29	4984	1704	2.92	1.027
30-34	5109	1619	3.16	1.018
35-39	3833	1170	3.28	0.997
40-44	2529	890	2.84	0.983
45-49	1824	775	2.35	1.002
50-54	1423	657	2.17	1.012
55-59	1197	638	1.88	1.060
60-64	965	536	1.80	1.036
65+	1791	1137	1.58	0.878

	n	% female	% under 15	% males 15-59	M:F ratio
names added in
2024-01-05 → 2024-03-29	7039	33.04	18.82	52.14	2.03
2024-03-29 → 2024-04-30	4377	26.37	18.14	56.18	2.79
2024-04-30 → 2024-06-30	7523	33.64	24.92	45.42	1.97
2024-06-30 → 2024-08-31	6658	31.86	24.89	48.23	2.14
2024-08-31 → 2025-03-23	18666	28.21	22.40	54.86	2.55
2025-03-23 → 2025-06-15	5244	30.19	23.61	53.03	2.31
2025-06-15 → 2025-07-15	3195	21.00	17.31	64.54	3.76
2025-07-15 → 2025-07-31	1826	26.29	21.85	56.52	2.80