Air Pollution - Is the Turning Point Real

Research question

Does the annual exposure series require a change in slope, or can a simpler trend explain it?

Assess whether an apparent change point in a short annual national PM2.5 exposure series is stable under alternative error assumptions, time windows and model choices. Compare a searched continuous hinge with a no-break linear trend and a smooth quadratic; distinguish break-year uncertainty conditional on a break from evidence that a break exists.

Data and comparison design

A hinge model joins two slopes at a candidate year. Searching many years makes the best-looking break easier to find by chance, so read search-adjusted tests alongside AICc and BIC, which penalize added complexity. A knot at the earliest eligible year deserves particular scrutiny.

Input or source	Role in this case	Source context
World Bank India PM2.5 API	Original national annual exposure response
World Bank indicator metadata	Units, source and indicator definition
IHME GBD 2023 source attribution	Upstream attribution from captured World Bank metadata; not downloaded by the study

Recorded findings

Model check	Independent errors	AR(1) errors
Selected hinge knot	1995	1995
Search-adjusted p-value	0.2060	0.3365
Sensitivity support	No AICc or BIC hinge wins in 48 settings	The settings are correlated, not 48 independent replications.

The 34 observations have mean 51.649594 and sample SD 2.032641 µg/m³. Requiring five observations strictly on each side gives 24 eligible knots, 1995–2018. Both searched hinges select 1995, an eligibility boundary.

Hinge-minus-line AICc/BIC differences are +1.438992/+3.148856 for independent errors and +2.310964/+3.631885 for AR(1), favoring the line. The no-break slopes remain positive, approximately +0.144895/+0.142971 µg/m³/year; no break does not mean no trend.

Among 1,999 null replicates per error model, 411 and 672 exceedances give add-one search-adjusted p-values 0.2060 and 0.3365. The hinge wins neither AICc nor BIC in any of 48 sensitivity settings. AICc line/quadratic counts are 42/6; BIC counts are 45/3. Half select boundary knots. Conditional knot intervals span 1995–2018, the full eligible range; these intervals assume a fitted hinge and do not establish break existence.

The selected series does not convincingly support a stable abrupt trend break under these comparisons. This is not proof that no real change occurred.

What the comparison shows

The apparent turning point loses support when search and model complexity are accounted for. A visually appealing bend therefore should not be presented as a demonstrated policy effect.

Assumptions and limits

No supported break does not mean no trend or prove that no real change occurred. National aggregation, short windows, exposure-model uncertainty and error assumptions constrain the finding. Conditional break-year intervals assume a break exists; correlated sensitivity settings are not independent votes.

Sources and research materials

The unchanged research package comes from the upstream archive. It retains the original conversation and included scientific artifacts, including analysis code, results and figures. The poster is an AI-generated thematic illustration, not study evidence. This introduction summarizes archived findings; the analysis was not rerun. Later bounded checks do not certify the entire study or mean the original package was corrected.