Tools

Four scripts, each from one methods note in the newsletter. Standard library only, Python 3.10 or later. python -m pytest tests/ runs the 23 tests.

Script	From	What it does
`deletion_rate.py`	Research praxis, July 2026	Reads an electoral-roll deletion figure: share of the roll, expected churn, draft or final, wrongful removals.
`poverty_lines.py`	Research praxis, May 2026	Puts World Bank and Indian committee poverty lines in rupees per person per month at one price year.
`odds_likelihood.py`	Methods corner, June 2024	Odds and probability; odds ratio against relative risk; a binomial likelihood on a grid.
`audits.py`	Research praxis, April 2025; methods note, June 2025	Two checklists as JSON: the four questions for a recommendation, the eight questions for a participatory design.

deletion_rate.py

python tools/deletion_rate.py --removed 4700000 --roll 78900000 --stage draft
python tools/deletion_rate.py --removed 4700000 --roll 78900000 \
    --years 22 --death-rate 6.4 --out-migration-rate 4 --stage draft

Prints the four checks. Share of the roll. Expected attrition, roll x (1 - (1 - r)^years), and the excess or shortfall against it. Whether the figure is a draft, and the restored share if you pass --final-removed. The number of wrongful removals if you pass an audited --wrongful-share.

There is no built-in attrition rate. Pass the crude death rate (Sample Registration System) and an out-migration rate (Census D-series or the PLFS migration module) per thousand per year.

poverty_lines.py

python tools/poverty_lines.py --year 2023
python tools/poverty_lines.py --check 1622 --line tendulkar_rural
python tools/poverty_lines.py --fetch

Ten lines in rupees per person per month: the World Bank lines at 2017 and 2021 PPPs, and the Tendulkar and Rangarajan lines, rural and urban. Dollar lines are converted with the PPP factor for private consumption in the base year, the CPI ratio to the target year, and 30.4375 days a month. Rupee lines are inflated by the CPI ratio.

--check finds the price year at which a line comes closest to a figure you have been given. --fetch pulls current PPP and CPI values from the World Bank API and prints what changed. The sources for every figure are in the docstring. This is for reading a table; it is not a poverty headcount.

odds_likelihood.py

python tools/odds_likelihood.py odds 0.25
python tools/odds_likelihood.py table 75 25 50 50
python tools/odds_likelihood.py lik 7 20 --points 11

odds and prob convert between the two. table takes a 2x2 table (exposed with the outcome, exposed without, unexposed with, unexposed without) and prints relative risk, odds ratio and their ratio. lik prints the binomial log-likelihood across values of p. No hypothesis test.

audits.py

python tools/audits.py recommendation --template > recs.json
python tools/audits.py recommendation recs.json
python tools/audits.py participation --template > design.json
python tools/audits.py participation design.json

Fill in the JSON, run the check. It exits 1 until every question is answered.

The recommendation audit asks four questions of each recommendation: which finding supports it, what assumption connects the two, what evidence supports the assumption, what would make it wrong. The participation audit asks eight questions about who holds authority, each answered researchers, shared or participants, and names the arrangement: consultation, collaboration or community control. Worked examples are in examples/.