Audit

Setiap angka yang terbit, diturunkan ulang oleh sesuatu yang bukan pipeline-nya.

Angka di ruang analisis ditulis sebagai jalur ke keluaran pipeline, jadi yang tampil selalu yang benar-benar diemitkan. Itu menjamin situs ini tidak salah kutip — tapi tidak membuktikan pipeline-nya benar. Program yang mengeluarkan angkanya sendiri paling jauh hanya konsisten dengan dirinya sendiri.

Maka ada notebook yang membaca berkas mentahnya dan keluaran yang terbit, menghitung ulang tiap klaim dari nol, lalu memastikan keduanya sama. Ia sengaja tidak mengimpor pipeline-nya: pipeline-nya ditulis dengan pustaka standar, notebook-nya dengan pandas dan numpy (pandas + numpy only, no scipy). Kalau dua implementasi yang tidak berbagi mesin sampai ke angka yang sama, angka itu bukan artefak salah satunya.

klaim diperiksa

pemeriksaan

gagal

Ia juga memeriksa tempat-tempat situs ini mengaku lemah, karena sebuah pengakuan juga sebuah klaim. Residual piece 01 yang bergantung skala dihitung ulang dengan dua cara, dan klaim kerataan piece 02 yang ditarik diperiksa memang layak ditarik — profil tidur benar-benar serata profil layar.

Setiap pemeriksaan

What Makes Writing Writing

16 pemeriksaan

the two layers →21 abilities →writers and authors →two raters →exposure and pay →

klaim	dihitung ulang	yang terbit
the two layers correlate	-0.367	-0.367	cocok
shared variance	0.135	0.135	cocok
CI low	-0.4658	-0.4658	cocok
CI high	-0.2591	-0.2591	cocok
predicted rank	267	267	cocok
rated rank	4	4	cocok
displacement	263	263	cocok
largest displacement of all	1	1	cocok
OLS slope is negative	-5.1569	-5.1569	cocok
residual rank, shipped	1	1	cocok
residual rank, OLS	17	17	cocok
admits scale dependence	True	True	cocok
one negative ability	1	1	cocok
it is Originality	'Originality'	'Originality'	cocok
its score	-0.144	-0.144	cocok
writers' percentile on Originality	97.4	97.4	cocok

The Line Someone Drew

17 pemeriksaan

screens against sleep →where the line falls →which ratios survive →every driver, item by item →

klaim	dihitung ulang	yang terbit
screen r	0.1344	0.1344	cocok
screen r2 %	1.8	1.8	cocok
sleep quality r2 %	14.5	14.5	cocok
variance ratio	8.0	8.0	cocok
spearman	0.1455	0.1455	cocok
absorbed by hours slept, %	46.0	46.0	cocok
exactly one cutoff reproduces the column	1	1	cocok
and it is	14	14	cocok
next best cutoff	13	13	cocok
next best agreement	0.9798	0.9798	cocok
grid size	130	130	cocok
lowest ratio	1.147	1.147	cocok
highest ratio	2.967	2.967	cocok
intervals clear of 1	92	92	cocok
screen spread ÷ size	0.218	0.218	cocok
sleep spread ÷ size	0.187	0.187	cocok
sleep is at least as flat	True	True	cocok

How Many People Is a Wage

45 pemeriksaan

the aggregate among the units →the 2021 column →the ratio, year by year →wage against local cost →twenty years of dispersion →a basket nobody buys →the floor, and the going rate →

klaim	dihitung ulang	yang terbit
rows in the province column	35	35	cocok
actual provinces	34	34	cocok
the aggregate is named	'INDONESIA'	'INDONESIA'	cocok
rho(ump2020, ump2022)	0.9958	0.9958	cocok
rho(ump2020, ump2021)	0.6812	0.6812	cocok
rho(ump2021, ump2022)	0.6723	0.6723	cocok
no province fell 2020 to 2022	True	True	cocok
provinces outside the [2020, 2022] bracket	13	13	cocok
displaced values traced to another province	9	9	cocok
clean 2021 provinces	21	21	cocok
of those, froze the wage exactly	18	18	cocok
rho after shifting the column back one row	0.2101	0.2101	cocok
the shift hypothesis is rejected	True	True	cocok
the row gaps are irregular	[-7, -2, -1, 1, 9, 15, 24]	[-7, -2, -1, 1, 9, 15, 24]	cocok
province-years in the clean panel	238	238	cocok
years used	[2015, 2016, 2017, 2018, 2019, 2020, 2022]	[2015, 2016, 2017, 2018, 2019, 2020, 2022]	cocok
median people, first year	4.674	4.674	cocok
median people, last year	4.774	4.774	cocok
fewest people, any province-year	2.792	2.792	cocok
most people, any province-year	8.861	8.861	cocok
share of province-years under five people	54.6	54.6	cocok
share under four people	8.0	8.0	cocok
one province is lowest in every clean year	True	True	cocok
elasticity, province fixed effects	1.1457	1.1457	cocok
elasticity, province + year fixed effects	-0.2621	-0.2621	cocok
the two-way interval contains zero	True	True	cocok
Theil T of provincial minimum wages, 2002	0.01982	0.01982	cocok
provinces with a wage in 2002	30	30	cocok
Theil T of provincial minimum wages, 2022	0.02086	0.02086	cocok
provinces with a wage in 2022	34	34	cocok
rank persistence of people-per-wage	0.9199	0.9199	cocok
provinces compared	34	34	cocok
food share of actual spending, 2022 PERKOTAAN	0.472	0.472	cocok
food share of the poverty line, 2022 PERKOTAAN	0.7278	0.7278	cocok
food share of actual spending, 2022 PERDESAAN	0.5652	0.5652	cocok
food share of the poverty line, 2022 PERDESAAN	0.7614	0.7614	cocok
median actual wage over legal minimum, 2015	1.3995	1.3995	cocok
provinces below their own legal minimum, 2015	0	0	cocok
median actual wage over legal minimum, 2022	1.1315	1.1315	cocok
provinces below their own legal minimum, 2022	9	9	cocok
every machine cell re-derives	0	0	cocok
machine cells committed	2376	2376	cocok
the default household is below the line	1	1	cocok
how many provinces put 4 people on 1 wage under the line	3	3	cocok
provinces excluded from the machine	['DKI JAKARTA']	['DKI JAKARTA']	cocok

The Denominator

20 pemeriksaan

a world median that barely moves →the countries moving underneath it →which half of the ratio the variance comes from →deaths of undetermined intent, by sex →the gap that reverses by age →a survey and a death register, disagreeing →

klaim	dihitung ulang	yang terbit
piece 04 · the world median in the first year	3.3523	3.3523	cocok
piece 04 · the world median in the last year	3.2915	3.2915	cocok
piece 04 · how far the yearly median ever wanders	0.2025	0.2025	cocok
piece 04 · median move, counted-death countries	0.7966	0.7966	cocok
piece 04 · median move, modelled countries	0.4051	0.4051	cocok
piece 04 · p-value that counted countries move more	0.0018	0.001794	cocok
piece 04 · correlation: ratio against the male rate	0.0482	0.0482	cocok
piece 04 · correlation: ratio against the female rate	-0.4843	-0.4843	cocok
piece 04 · share of the ratio's variance from the female side	104.1	104.1	cocok
piece 04 · share of the ratio's variance from the male side	-4.1	-4.1	cocok
piece 04 · countries in the undetermined-intent comparison	97	87	cocok
piece 04 · median share filed as undetermined, women	0.2505	0.2542	cocok
piece 04 · median share filed as undetermined, men	0.2059	0.2097	cocok
piece 04 · share of countries where women's undetermined-share is higher	0.71	0.7356	cocok
piece 04 · p-value, women's share higher than men's	8.5e-05	1.279e-05	cocok
piece 04 · undetermined-share gap at 15-19 (women minus men)	-0.081	-0.0806	cocok
piece 04 · undetermined share at 95-99, women	0.723	0.7227	cocok
piece 04 · undetermined share at 95-99, men	0.397	0.3973	cocok
piece 04 · symptom ratio, women:men (US household survey)	1.231	1.231	cocok
piece 04 · suicide-death ratio, men:women (US death register)	3.69	3.688	cocok

analysis/notebooks/audit.ipynb

Seluruh notebook-nya di bawah — 52 sel, dengan keluaran yang benar-benar dihasilkan tiap sel waktu analysis/notebooks/make_audit.py + analysis/04-gender-paradox/make_audit_04.py menjalankannya. Tidak ada yang disalin tangan, dan tidak ada satu baris pun yang berupa tangkapan layar.