Tests

The record, put to statistical test

Hypothesis tests on the data this site publishes and on public datasets from the UN, the Palestinian Central Bureau of Statistics, the Israel Prison Service, B’Tselem, Yesh Din and the Committee to Protect Journalists: who is killed, the ceasefires, the ICJ orders, the West Bank, the named list of the dead, settler impunity, Area C, food prices, detention without charge, journalists and the destruction of buildings. Each test is stated with its null hypothesis, its data, its effect size and its confidence interval, and every figure on this page is read out of the published data file.

25

Tests

each with its null hypothesis, data, effect size and interval

73k

Identity numbers checked

every one passes the Population Registry check digit

20

Nulls rejected

after correction for multiple testing

3

ICJ orders with no detectable fall

in the daily death rate after the order

In plain terms

What the tests show

The named list of the dead is genuine on every internal test. Every one of the 72,835 identity numbers on the Ministry of Health list passes the Population Registry check digit, which a fabricated number passes about one time in ten; none is duplicated; and every stated age agrees with its date of birth. H6a

The dead are not confined to men of fighting age. Just over half of the identified dead are women, children or people aged 60 and over. H1b

Every ceasefire cut the rate of killing, and none stopped it. Killings continued through the November 2023 pause, the January to March 2025 ceasefire and the ceasefire in force since October 2025. H2c

The three ICJ provisional measures orders of 2024 were not followed by any detectable fall in the daily death rate. H3a

In the West Bank, Palestinians were killed at more than four times the earlier rate after 7 October 2023, and settler incidents in 2024 ran at about three and a half times the 2022 rate. H4a

No state that withholds recognition of Palestine has a government that has called the conduct in Gaza genocide; 50 recognising states do. H8b

Settlers who attack Palestinians are rarely convicted. Only 3% of the police files Yesh Din followed ended in a conviction, and an indicted settler is convicted about half the time, against 99.7% for Palestinians in the military courts. H9a

In Area C, under 1% of Palestinian building applications are approved, and an unpermitted Palestinian structure is about two and a half times as likely to be demolished as a settler's. H10b

Food prices rose by about 30% during the total blockade of 2025, and across the war food has cost about three times its pre-war price. H11a

Administrative detention without charge is nearly eight times its 2008-2022 level. H13

Journalists have been killed at more than four times the rate of the Iraq war and seven times that of Syria's, on CPJ's lowest count. H14a

Destroyed buildings rose at every satellite assessment, and about 11,000 more were recorded destroyed after the October 2025 ceasefire. H15b

In the 2008-09 and 2014 Gaza wars, most of those killed took no part in the hostilities, on B'Tselem's case-by-case count. H16c

The rates

Every rate ratio on one axis

The ceasefires, the ICJ orders, the West Bank, food prices under blockade, detention without charge and journalists killed, each as a ratio with its 95% interval. A ratio of 1 is no change. Green is a fall that survives the correction, red a rise, grey no detectable change.

Ratios with 95% intervals

Log scale. Each line is the interval; the dot is the estimate.

data/tests.json

Interactive chart — open it in the live record.

H1

Who is killed: age and sex against the population

H1b

Age and sex of the dead: pattern confined to men of fighting age

null rejected

52.1% of the identified dead (95% CI 51.7% to 52.5%) are women, children under 18 or people aged 60 and over, against 75.5% of the population. More than half of the dead fall outside the group from which armed men are drawn: the dead are not confined to men of fighting age.

Null hypothesis: The identified dead are confined to men aged 18-59.
Data: Ministry of Health named list, 72,835 identified dead; PCBS/UNFPA Gaza Strip population 2023 (2,226,544).
Test: Exact binomial test of the share of women, children and people aged 60 and over against zero
Effect: Share of dead who are women, children under 18 or aged 60+: 0.521 (95% CI 0.517 to 0.525)
p: < 0.001 · adjusted for false discovery < 0.001
Report: §6.1

H2

The ceasefires

Daily deaths in Gaza, and new killings during each ceasefire

Grey is the daily increase in the reported total. Inside each shaded ceasefire, green is new killings alone, with bodies recovered from the rubble taken out. Log scale; drag the slider to zoom.

data/tests.json · H2

Interactive chart — open it in the live record.

H2a

The November 2023 pause and the daily killing rate

null rejected

Reported killings fell from 284.4 a day in the fortnight before the pause to 0.71 a day during it (rate ratio 0.0025, 95% CI 0.0013 to 0.0047). Killing did not stop: at least 5 people were reported killed in the 7 days. Against the upper bound of 40 deaths the ratio is 0.0201, still a fall of about 98%.

Null hypothesis

The daily rate of killing during the pause of 24-30 November 2023 equalled the rate in the 14 days before it.

Data

Comparison window: Ministry of Health / Government Media Office daily series (Tech For Palestine), 10-23 November 2023. Pause: killings reported incident by incident (OCHA Flash Updates #49-#55; AP; Al Jazeera).

Test

Negative binomial regression on a pause indicator, Newey-West errors (lag 7)

Effect

Rate ratio, pause against the 14 days before: 0.00250 (95% CI 0.00135 to 0.00468)

p

< 0.001 · adjusted for false discovery < 0.001

Sensitivity

Exact conditional Poisson test: 0.00250 (95% CI 0.000815 to 0.00587), p < 0.001
Against the upper bound of 40 deaths during the pause: 0.0201 (95% CI 0.0143 to 0.0274), p < 0.001

Sources

Hostilities in the Gaza Strip and Israel - Flash Update #49 — OCHA, 24 November 2023
Hostilities in the Gaza Strip and Israel - Flash Update #51 — OCHA, 26 November 2023
Hostilities in the Gaza Strip and Israel - Flash Update #53 — OCHA, 28 November 2023
Hostilities in the Gaza Strip and Israel - Flash Update #55 — OCHA, 30 November 2023
Israeli troops fire at Palestinians attempting to return to northern Gaza during cease-fire — Associated Press (PBS NewsHour), 24 November 2023
'The war is not over': Israel blocks Palestinians' return to northern Gaza — Al Jazeera, 24 November 2023

Report

§13

H2b

The January-March 2025 ceasefire and the daily killing rate

null rejected

New killings fell from 48.8 a day in the 60 days before to 7.28 a day during the ceasefire (rate ratio 0.149, 95% CI 0.051 to 0.44). Killing continued throughout: 422 new killings were reported in 58 days (95% CI 6.6 to 8.0 a day).

Null hypothesis

The daily rate of new killings during the ceasefire (19 Jan 2025 to 17 Mar 2025) equalled the rate in the 60 days before it.

Data

Ministry of Health daily series (Tech For Palestine). Ceasefire days count new killings only where the Ministry separates them from bodies recovered.

Test

Negative binomial regression on a ceasefire indicator, Newey-West errors (lag 7)

Effect

Rate ratio, ceasefire against the 60 days before: 0.149 (95% CI 0.0506 to 0.440)

p

< 0.001 · adjusted for false discovery < 0.001

Sensitivity

Weekly totals: 0.152 (95% CI 0.0654 to 0.355), p < 0.001
New killings only in both windows: 0.180 (95% CI 0.0652 to 0.496), p < 0.001

Report

§13

H2c

The October 2025 ceasefire and the daily killing rate

null rejected

New killings fell from 96.1 a day in the 60 days before to 3.74 a day during the ceasefire (rate ratio 0.039, 95% CI 0.027 to 0.055). Killing continued throughout: 1,336 new killings were reported in 357 days (95% CI 3.54 to 3.95 a day).

Null hypothesis

The daily rate of new killings during the ceasefire (10 Oct 2025 to 01 Oct 2026) equalled the rate in the 60 days before it.

Data

Ministry of Health daily series (Tech For Palestine). Ceasefire days count new killings only where the Ministry separates them from bodies recovered.

Test

Negative binomial regression on a ceasefire indicator, Newey-West errors (lag 7)

Effect

Rate ratio, ceasefire against the 60 days before: 0.0390 (95% CI 0.0274 to 0.0553)

p

< 0.001 · adjusted for false discovery < 0.001

Sensitivity

Weekly totals: 0.0370 (95% CI 0.0234 to 0.0576), p < 0.001
New killings only in both windows: 0.0610 (95% CI 0.0465 to 0.0800), p < 0.001

Report

§13

H3

The ICJ provisional measures orders

Daily deaths and the three ICJ provisional measures orders

Grey is daily deaths, December 2023 to July 2024. Blue is the segmented model fitted to them, which allows a change in level and in trend at each order. No step down follows any of the three.

data/tests.json · H3

Interactive chart — open it in the live record.

H3a

The ICJ order of 26 January 2024 and the daily killing rate

null not rejected

No detectable change in the level or the trend of daily deaths at the order (level rate ratio 1.006, 95% CI 0.81 to 1.3; joint test p = 0.99). The order was not followed by a measurable fall in the rate of killing. This is the absence of a detectable change, not proof that the order had no effect on any conduct.

Null hypothesis: No change in the level or the slope of daily deaths at the order of 26 January 2024.
Data: Ministry of Health daily series (Tech For Palestine), 1 December 2023 to 31 July 2024.
Test: Segmented negative binomial regression (trend, and a level and slope change at each order), Newey-West errors; joint Wald test of level and slope
Effect: Rate ratio, level change at the order: 1.01 (95% CI 0.808 to 1.25)
p: 0.989 · adjusted for false discovery 0.989
Report: §13

H3b

The ICJ order of 28 March 2024 and the daily killing rate

null not rejected

No detectable change in the level or the trend of daily deaths at the order (level rate ratio 0.819, 95% CI 0.65 to 1; joint test p = 0.16). The order was not followed by a measurable fall in the rate of killing. This is the absence of a detectable change, not proof that the order had no effect on any conduct.

Null hypothesis: No change in the level or the slope of daily deaths at the order of 28 March 2024.
Data: Ministry of Health daily series (Tech For Palestine), 1 December 2023 to 31 July 2024.
Test: Segmented negative binomial regression (trend, and a level and slope change at each order), Newey-West errors; joint Wald test of level and slope
Effect: Rate ratio, level change at the order: 0.819 (95% CI 0.645 to 1.04)
p: 0.157 · adjusted for false discovery 0.175
Report: §13

H3c

The ICJ order of 24 May 2024 and the daily killing rate

null not rejected

No detectable change in the level or the trend of daily deaths at the order (level rate ratio 1.115, 95% CI 0.7 to 1.8; joint test p = 0.81). The order was not followed by a measurable fall in the rate of killing. This is the absence of a detectable change, not proof that the order had no effect on any conduct.

Null hypothesis: No change in the level or the slope of daily deaths at the order of 24 May 2024.
Data: Ministry of Health daily series (Tech For Palestine), 1 December 2023 to 31 July 2024.
Test: Segmented negative binomial regression (trend, and a level and slope change at each order), Newey-West errors; joint Wald test of level and slope
Effect: Rate ratio, level change at the order: 1.11 (95% CI 0.703 to 1.77)
p: 0.812 · adjusted for false discovery 0.840
Report: §13

H4

The West Bank before and after 7 October 2023

H4a

West Bank killings before and after 7 October 2023

null rejected

Palestinians were killed in the West Bank at 3.48 a day after 7 October 2023 against 0.75 before it, 4.66 times the rate (95% CI 3.89 to 5.59), in a year that was already the deadliest on record before the war began.

Null hypothesis: The daily rate of Palestinians killed in the West Bank was the same after 7 October 2023 as before it, within 2023.
Data: OCHA via the report: 208 killed in 279 days (1 January-6 October 2023); 299 in 86 days (7 October-31 December 2023).
Test: Exact conditional test of two Poisson rates
Effect: Rate ratio, after against before: 4.66 (95% CI 3.89 to 5.59)
p: < 0.001 · adjusted for false discovery < 0.001
Report: §11

H4b

Settler incidents against Palestinians, 2024 against 2022

null rejected

Settler incidents ran at 3.5 times the 2022 rate in 2024 (95% CI 3.22 to 3.81), and at 2.6 times even at the rounding bounds least favourable to the finding.

Null hypothesis

The rate of settler incidents against Palestinians was the same in 2024 as in 2022.

Data

OCHA daily averages via the report: 2 a day in 2022, 7 a day in 2024.

Test

Exact conditional test of two Poisson rates on annual counts reconstructed from the daily averages

Effect

Rate ratio, 2024 against 2022: 3.50 (95% CI 3.22 to 3.81)

p

< 0.001 · adjusted for false discovery < 0.001

Sensitivity

At the rounding bounds least favourable to rejection (2.5 and 6.5 a day): 2.60 (95% CI 2.41 to 2.81), p < 0.001

Report

§11

H6

Is the named list genuine?

Single-year ages on the named list of the dead

Ages ending in 0 or 5 in red. A list built from guesses piles up on those ages; this one shows no heaping, because the ages follow from registry dates of birth.

data/tests.json · H6

Interactive chart — open it in the live record.

H6a

The identity numbers on the named list

validity not rejected; the 10% fabrication benchmark is rejected

All 72,835 of the 72,835 identity numbers pass the Population Registry check digit, where fabricated numbers would pass about one time in ten. There are 0 duplicate identity numbers and 0 duplicate name-and-birth-date pairs, and the stated age agrees with the date of birth in 72,835 of 72,835 records. The list as published is consistent with registry data. These tests cannot detect a real identity number attached to someone who did not die; that question needs matching against other records.

Null hypothesis: The identity numbers are valid Population Registry numbers (pass the check digit).
Data: Ministry of Health named list, 72,835 records.
Test: Share passing the check digit, with its interval; exact binomial test against the 10% expected of fabricated numbers; decision rule: validity rejected if under 99% pass
Effect: Share of identity numbers with a valid check digit: 1.00 (95% CI 1.00 to 1.00)
p: < 0.001 · adjusted for false discovery < 0.001
Report: §6.1

H6b

Age heaping on the named list

highly accurate

No age heaping: Whipple's index is 100.0 and Myers' index 0.85, against the 2017 census's 100 and 2.4. Ages on the list follow from registry dates of birth rather than estimates, which is what a list built from identity records, and not from guesses, should show.

Null hypothesis: Age reporting shows no more heaping on ages ending in 0 and 5 than a registry-based list would.
Data: As H6a.
Test: Whipple's index (ages 23-62) on the UN scale, and Myers' blended index (ages 10-89), against the PCBS 2017 census
Effect: Whipple's index: 100
p: —
Report: §6.1

H8

What states have done

H8b

Recognition of Palestine and calling it genocide

null rejected

50 of 157 recognising states have a government that has called it genocide; none of the 36 non-recognising states has (odds ratio 34.29, with a correction for the empty cell).

Null hypothesis: Whether a UN member state government has called it genocide is independent of whether it recognises Palestine.
Data: World positions dataset, 193 UN member states.
Test: Fisher's exact test (interval with Haldane correction for a zero cell)
Effect: Odds ratio: 34.3 (95% CI 2.06 to 570)
p: < 0.001 · adjusted for false discovery < 0.001
Report: §15.4

H9

Settler violence and the courts

H9a

Settlers and Palestinians before Israeli courts in the West Bank

null rejected

Of Israelis indicted for offences against Palestinians in the West Bank, 49% were convicted in full or in part (95% CI 39% to 58%); of Palestinians tried in the military courts, 99.74% were. Only 3% of the police files Yesh Din followed ended in any conviction. The two figures come from different years and different courts, and a partial conviction is counted as a conviction on both sides.

Null hypothesis

An Israeli indicted for an offence against Palestinians in the West Bank is as likely to be convicted as a Palestinian tried in the military courts.

Data

Yesh Din, 1,701 police files on offences by Israelis against Palestinians, 2005-2024 (109 indictments, 53 full or partial convictions); Israeli military courts' 2010 annual report as reported by Haaretz (9,542 cases, 25 full acquittals).

Test

Fisher's exact test of two proportions

Effect

Odds ratio, settler conviction against Palestinian conviction: 0.00250 (95% CI 0.00144 to 0.00428)

p

< 0.001 · adjusted for false discovery < 0.001

Sources

Data Sheet: Law Enforcement on Israeli Civilians in the West Bank, 2005-2024 — Yesh Din, 16 January 2025
Nearly 100% of all military court cases in West Bank end in conviction — Haaretz, 29 November 2011

Report

§11

H9b

Palestinian victims of settler violence who do not complain to the police

null rejected

In 66% of the attacks Yesh Din documented in 2024, the Palestinian victims chose not to complain to the Israeli police, most citing distrust of the investigation or fear of reprisal. The prosecution figures in H9a therefore start from a minority of the attacks.

Null hypothesis: No more than half of Palestinian victims of settler violence decline to file a police complaint.
Data: Yesh Din, incidents it documented in 2024: in 101 of 153 the victims chose not to complain.
Test: Exact binomial test against one half (one-sided)
Effect: Share of victims not filing a complaint: 0.660 (95% CI 0.592 to 1.00)
p: < 0.001 · adjusted for false discovery < 0.001
Report: §11

H10

Building and demolition in Area C

H10a

Building permits for Palestinians in Area C

null rejected

Between 2016 and 2020, 24 of 2,550 Palestinian applications for building permits in Area C were approved: 0.9% (95% CI up to 1.3%). The Civil Administration told the Knesset that 60-70% of settlement plans are approved. The units differ (permits against plans), but the gap is more than sixty-fold.

Null hypothesis

Palestinian applications for building permits in Area C are approved at the rate settlers' plans are approved (at least 60%).

Data

Civil Administration and Bimkom figures, 2016-2020: 24 of 2,550 Palestinian applications approved; the Civil Administration's own statement that 60-70% of settlement plans submitted are approved (Knesset, 19 July 2023).

Test

Exact binomial test against the lower bound of the settler approval rate (one-sided)

Effect

Share of Palestinian applications approved: 0.00940 (95% CI 0.00 to 0.0132)

p

< 0.001 · adjusted for false discovery < 0.001

Sources

The Civil Administration acknowledges extreme discrimination in building permits and law enforcement between Palestinians and settlers — Peace Now, 10 August 2023

Report

§12

H10b

Demolition of unpermitted structures: Palestinian and settler

null rejected

Of structures the Civil Administration identified as built without a permit in 2022 and the first half of 2023, 30% of Palestinian structures were demolished against 14% of settler structures: the odds of demolition were 2.48 times higher for a Palestinian structure (95% CI 1.9 to 3.2).

Null hypothesis: An unpermitted Palestinian structure in Area C is as likely to be demolished as an unpermitted settler structure.
Data: Civil Administration figures, 2022 to mid-2023: 770 of 2,600 Palestinian structures identified as unpermitted demolished, against 88 of 606 settler structures.
Test: Fisher's exact test of two proportions
Effect: Odds ratio, Palestinian against settler: 2.48 (95% CI 1.95 to 3.15)
p: < 0.001 · adjusted for false discovery < 0.001
Report: §12

H11

Food prices in Gaza

Food prices in Gaza against their pre-war level

The geometric mean of each food item’s monthly price divided by its price before 7 October 2023. PCBS, November 2023 to August 2026.

data/tests.json · H11

Interactive chart — open it in the live record.

H11a

The total blockade of March to May 2025 and food prices in Gaza

null rejected

During the total blockade of March to May 2025, food prices in Gaza rose 30% above the three months before (95% CI 13% to 50%), from 3.89 to 4.88 times their pre-war level.

Null hypothesis

Food prices in the total blockade (March-May 2025) were the same as in the three months before (December 2024-February 2025).

Data

PCBS monthly prices in Gaza of 32 food items with a complete series (COICOP division 01), November 2023 to August 2026.

Test

Regression of log price on a period indicator with item fixed effects, errors clustered by item

Effect

Price ratio, after against before: 1.30 (95% CI 1.13 to 1.50)

p

< 0.001 · adjusted for false discovery < 0.001

Sources

State of Palestine: Prices of Basic Commodities in Gaza — Palestinian Central Bureau of Statistics (HDX), updated 13 September 2026

Report

§6.3

H11c

Food prices in Gaza against their pre-war level

null rejected

Across the 34 months of data, food in Gaza cost on average 3.16 times its pre-war price (95% CI 2.6 to 3.9), and was above its pre-war price in 34 of 34 months. At the peak, in 2025-06, it cost 11.8 times as much.

Null hypothesis: Food prices in Gaza during the war were at their pre-war level.
Data: As H11a; each month's index is the geometric mean of the 32 items' prices relative to their average before 7 October 2023.
Test: One-sample t-test of the monthly log index against zero
Effect: Mean price level against pre-war (geometric): 3.16 (95% CI 2.60 to 3.85)
p: < 0.001 · adjusted for false discovery < 0.001
Report: §6.3

H13

Detention without charge

Palestinians held without charge or trial, at each year end

Administrative detainees and those held under the Unlawful Combatants Law, Israel Prison Service figures compiled by B’Tselem. 2026 is 30 June.

data/tests.json · H13

Interactive chart — open it in the live record.

H13

Administrative detention, without charge or trial, before and after 7 October 2023

null rejected

Israel held on average 3,283 Palestinians in administrative detention, without charge or trial, at each year end since October 2023, against 418 in 2008-2022: 7.85 times as many (95% CI 6.17 to 9.99). Every figure since 2023 is more than four times the highest year before it (784). These figures exclude those held as 'unlawful combatants'.

Null hypothesis

The number of Palestinians held in administrative detention was the same after October 2023 as in 2008-2022.

Data

Israel Prison Service figures compiled by B'Tselem, end of each year 2008-2025 and 30 June 2026.

Test

Negative binomial regression of the yearly count on a post-October-2023 indicator, Newey-West errors (lag 1)

Effect

Ratio, after against 2008-2022: 7.85 (95% CI 6.17 to 9.99)

p

< 0.001 · adjusted for false discovery < 0.001

Sources

Statistics on Palestinians in Israeli custody — B'Tselem, June 2026

Report

§11

H14

Journalists killed

Journalists killed a year, by war

Committee to Protect Journalists. The Gaza figure is CPJ’s count after its June 2026 review, the lowest it has published.

data/tests.json · H14

Interactive chart — open it in the live record.

H14a

Journalists killed a year: the Gaza war against the Iraq war

null rejected

Journalists were killed by Israel in the Gaza war at 76.6 a year, against 17.2 a year in the Iraq war: 4.46 times the rate (95% CI 3.6 to 5.54). The Gaza figure is the lowest CPJ has published, after its June 2026 review removed names.

Null hypothesis

Journalists were killed at the same rate in the Gaza war as in the Iraq war.

Data

Committee to Protect Journalists: 209 journalists killed by Israel in the Gaza war to 30 June 2026 (CPJ's count after its June 2026 review, the lowest it has published) over 2.73 years; Iraq war: as counted by CPJ.

Test

Exact conditional test of two Poisson rates

Effect

Rate ratio, Gaza war against the Iraq war: 4.46 (95% CI 3.60 to 5.54)

p

< 0.001 · adjusted for false discovery < 0.001

Sources

Israel-Gaza war — Committee to Protect Journalists, 2026
Iraq war and news media: a look inside the death toll — Committee to Protect Journalists, March 2013

Report

§10

H14b

Journalists killed a year: the Gaza war against the Syrian civil war

null rejected

Journalists were killed by Israel in the Gaza war at 76.6 a year, against 10.3 a year in the Syrian civil war: 7.46 times the rate (95% CI 6 to 9.3). The Gaza figure is the lowest CPJ has published, after its June 2026 review removed names.

Null hypothesis

Journalists were killed at the same rate in the Gaza war as in the Syrian civil war.

Data

Committee to Protect Journalists: 209 journalists killed by Israel in the Gaza war to 30 June 2026 (CPJ's count after its June 2026 review, the lowest it has published) over 2.73 years; Syrian civil war: as counted by CPJ.

Test

Exact conditional test of two Poisson rates

Effect

Rate ratio, Gaza war against the Syrian civil war: 7.46 (95% CI 6.00 to 9.30)

p

< 0.001 · adjusted for false discovery < 0.001

Sources

Israel-Gaza war — Committee to Protect Journalists, 2026
Iraq war and news media: a look inside the death toll — Committee to Protect Journalists, March 2013

Report

§10

H15

The destruction of buildings

Structures destroyed in Gaza at each satellite assessment

UNOSAT comprehensive damage assessments, by imagery date. “Damaged or destroyed” includes the “possibly damaged” category UNOSAT added in May 2024.

data/tests.json · H15

Interactive chart — open it in the live record.

H15a

Destroyed buildings in Gaza across twelve satellite assessments

null rejected

Destroyed structures rose at every one of the twelve UNOSAT assessments, from 6,583 in November 2023 to 134,422 in June 2026 (Kendall's tau 1.0). The destruction was not a single episode; it continued through every phase of the war.

Null hypothesis

The number of destroyed structures shows no monotonic trend across the assessments.

Data

UNOSAT comprehensive damage assessments, 12 imagery dates from 2023-11-07 to 2026-06-16.

Test

Kendall's tau (Mann-Kendall trend test)

Effect

Kendall's tau: 1.00

p

< 0.001 · adjusted for false discovery < 0.001

Sources

Gaza Strip Comprehensive Damage Assessment — UNOSAT, 2023-2026

Report

§6.4

H15b

Buildings destroyed during the October 2025 ceasefire

run

10,958 more structures were recorded as destroyed between the October 2025 ceasefire and June 2026, about 44 a day for 248 days. Some will be buildings damaged earlier that later collapsed or were cleared; the satellite count cannot separate those from new demolitions.

Null hypothesis: No further structures were destroyed in Gaza between the ceasefire of October 2025 and June 2026.
Data: UNOSAT: 123,464 structures destroyed at 11 October 2025, 134,422 at 16 June 2026.
Test: Count of newly destroyed structures, with an exact Poisson interval for the daily rate
Effect: Structures destroyed a day during the ceasefire: 44.2 (95% CI 43.4 to 45.0)
p: —
Report: §6.4

H16

Earlier Gaza wars: who was killed

H16a

Palestinians killed who took no part in hostilities: Cast Lead, 2008-09

null rejected

Of those killed whose status B'Tselem could determine, 57% took no part in the hostilities (95% CI from 55%). This counts the police officers killed on the first day as taking part, the reading least favourable to the finding.

Null hypothesis

No more than half of the Palestinians killed by Israeli forces, of those whose status B'Tselem could determine, took no part in the hostilities.

Data

B'Tselem: 773 took no part, 330 took part, 248 police officers (counted here as taking part), 36 undetermined (left out).

Test

Exact binomial test against one half (one-sided)

Effect

Share who took no part in hostilities: 0.572 (95% CI 0.550 to 1.00)

p

< 0.001 · adjusted for false discovery < 0.001

Sensitivity

Police officers left out rather than counted as taking part: 0.701 (95% CI 0.677 to 1.00), p < 0.001

Sources

Fatality figures — B'Tselem,

Report

§10

H16c

Palestinians killed who took no part in hostilities: Protective Edge, 2014

null rejected

Of those killed whose status B'Tselem could determine, 64% took no part in the hostilities (95% CI from 63%).

Null hypothesis: No more than half of the Palestinians killed by Israeli forces, of those whose status B'Tselem could determine, took no part in the hostilities.
Data: B'Tselem: 1,391 took no part, 765 took part, 46 undetermined (left out).
Test: Exact binomial test against one half (one-sided)
Effect: Share who took no part in hostilities: 0.645 (95% CI 0.628 to 1.00)
p: < 0.001 · adjusted for false discovery < 0.001
Report: §10

Method

How the tests were run

Tests: Counts are modelled as negative binomial, after a likelihood-ratio test rejected the Poisson model in every series where it was applied. Time series use Newey-West standard errors with a lag of seven days. Rates are compared with the exact conditional test for two Poisson rates. Two-by-two tables use Fisher's exact test.
Multiple testing: p-values are adjusted for the false discovery rate (Benjamini-Hochberg, 5%) across all the 30 tests with a p-value. Both raw and adjusted p-values are reported.
Caution: A change in a series at a dated event is an association, not proof that the event caused it. Tests on the named list inherit any bias in who has been identified.
What the tests are not: A test of a ceasefire or a court order measures whether the death rate changed around a date, not why. A test on the named list measures the list as published, not the deaths it does not contain: the 1,205 counted dead who have not been identified, and the dead still under the rubble, are outside it.
Limits: The tests are only as good as the data. The Ministry of Health's daily series changed source during the war (to the Government Media Office from 11 November 2023, and back), carries days on which no report was issued, and, from 2025, separates new killings from bodies recovered only on some days. The named list covers the identified dead only. The population baseline is the PCBS projection for 2023, before the war. Where a test depends on one of these weaknesses, its entry says so, and a sensitivity analysis tests whether the result survives it.

data/tests.json · generated 2026-10-02.