Protocol
v1.1.1 · frozen
DAH-PM25-2026-27-v1-1-1
Research · prospective observational study
Primary study activePersistent site-level contrasts in hourly PM2.5 across Delhi-NCR relative to the concurrent network median: a prospective observational study
The primary analysis asks whether physical monitoring sites show persistent hourly departures from the pollution level seen across the Delhi-NCR network at the same time.
External registration
Registered under embargo
OSF Preregistration
Primary period
1 Oct 2026
through 31 Mar 2027 · IST
Primary outcome
Hourly PM2.5
µg/m³ · physical site × hour
Primary comparison
A site is compared with the network at the same hour.
This separates a location-specific contrast from pollution changes that affect much of Delhi-NCR at once. The network median is a statistical reference, not an official city average or a personal-exposure estimate.
Rit = PM2.5it − medianj(PM2.5jt)
Interpretation
What a residual means
1
Site-hour value
Take PM2.5 at one eligible physical monitoring site in one eligible hour.
2
Concurrent network reference
Calculate the median PM2.5 across eligible physical sites reporting in that same hour.
3
Network-relative residual
Subtract the network median. Positive values are above the concurrent network reference; negative values are below it.
Study timeline
Pilot work is kept separate from the primary study.
September data are used for pipeline development, quality-control checks and descriptive review. They are excluded from the prespecified primary analysis.
Current
Pilot / method development
6–30 Sep 2026
Source checks, QC development and exploratory descriptions. Excluded from the primary estimand.
Prespecified
Primary collection window
1 Oct 2026–31 Mar 2027
Eligible physical-site × hour observations contribute under the frozen protocol rules.
Planned
Primary report
After sufficient primary data
Primary uncertainty summaries begin after at least 30 represented primary-period days; the study window still ends on 31 Mar 2027.
Primary-study readiness
What must be locked before 1 October.
A credible prospective study is not just a long archive. The question, cohort, QC, analysis code, registration status and provenance boundary must be explicit before primary-period data are interpreted.
Research question & estimand
Protocol v1.1.1 fixes the primary network-relative residual and site-effect summary.
Study window
Pilot ends 30 Sep; primary observation window is 1 Oct 2026–31 Mar 2027.
External registration
The website reports the registry state directly and must never imply OSF registration before it exists.
Baseline physical-site cohort
Frozen at 1 Oct 2026, 12:30 am with 73 eligible sites.
Missingness & QC
No imputation; critical observation-specific QC exclusions only; review flags remain unless the normalized value is invalid.
Frozen reproducible export
Formal results should cite an immutable release/export, analysis-spec version, code version and checksums.
Primary-analysis integrity gate
Primary interpretation is blocked by: EXTERNAL_REGISTRATION.
Operational watchdog flags do not automatically become analytical exclusions. Any deviation after the freeze belongs in an append-only deviation register rather than a silent code change.
How this connects to the public tools
The same provenance-tracked archive serves two different jobs.
Campus pages, Compare and History use the archive for transparent public exploration. The research layer applies separately versioned eligibility, QC and analysis rules. A public comparison is therefore not automatically a primary-study result.
Pilot analyses · not primary-study results
Live descriptive checks from the growing archive
These analyses help verify the pipeline and describe what has been collected so far. They should not be interpreted as the confirmatory result of the October–March study.
Evidence maturity
Short-term archive · 15 calendar days
Patterns are emerging, but remain sensitive to the particular days represented.
Delhi feeds
69
Local hours
277
PM2.5 monitor-hours
14,643
Latest reporting
57
Window: 24 Sept 2026, 5:00 am → 24 Oct 2026, 4:00 amTimezone: Asia/KolkataAnalysis spec 1.2.0Methods 1.2.0
Diurnal PM2.5 · Delhi local clock
What does the pilot archive look like by hour of day?
For each observed local hour, the analysis first takes the median PM2.5 across reporting Delhi station feeds. Those network-hour medians are then grouped by local clock hour across calendar days.
Clock hours repeated on ≥2 days
22/24
Q25–Q75 shown only when that clock hour has ≥2 observed daysMedian of Delhi network-hour PM2.5 medians
Current interpretation
The current archive is too short to estimate a stable daily cycle. Most clock hours currently represent only one or two calendar days. The figure is useful for checking the analysis pipeline and describing the data collected so far, but it should not be used to infer a stable morning or evening pattern.
Station-to-station correlation · exploratory
How closely are Delhi monitors moving together?
Pearson correlation is calculated only over local-hour PM2.5 buckets observed at both station feeds. Same-coordinate feeds are excluded from the distribution summary.
Independent pairs · n ≥ 12
2,054
25th percentile r
0.68
Median r
0.86
75th percentile r
0.92
Interpretation boundary
15 calendar days are represented. The high correlations mostly describe common movement during this short window. They are not yet evidence of persistent spatial structure. Short-window negative correlations can also be unstable.
Highest observed r
Pusa, Delhi - DPCC × Talkatora Garden, Delhi - DPCC
r 0.99270 shared local-hour PM2.5 buckets
IIT Delhi, Delhi - IITM × JNU, Delhi - DPCC
r 0.99233 shared local-hour PM2.5 buckets
Commonwealth Sports Complex, Delhi - DPCC × Major Dhyan Chand National Stadium, Delhi - DPCC
r 0.98269 shared local-hour PM2.5 buckets
IGI Airport (T3), Delhi - IITM × Jharoda Kalan, Delhi-DPCC
r 0.98214 shared local-hour PM2.5 buckets
Alipur, Delhi - DPCC × Narela, Delhi - DPCC
r 0.98272 shared local-hour PM2.5 buckets
Lowest observed r
Cantonment Area, Delhi - DPCC × NSIT Dwarka, Delhi - CPCB
r -0.15261 shared local-hour PM2.5 buckets
IGNOU_Maidan Garhi, Delhi - DPCC × NSIT Dwarka, Delhi - CPCB
r 0.11261 shared local-hour PM2.5 buckets
Ghoga ,Delhi-DPCC × Vasundhara Enclave, Delhi - DPCC
r 0.13231 shared local-hour PM2.5 buckets
Ghoga ,Delhi-DPCC × Karala,Delhi-DPCC
r 0.14217 shared local-hour PM2.5 buckets
Cantonment Area, Delhi - DPCC × Karala,Delhi-DPCC
r 0.16230 shared local-hour PM2.5 buckets
Named pair examples are a display-only exploratory feature, not part of the prospective primary study. Once at least 7 calendar days are represented, they additionally require n ≥ 207 shared hours (max of 24 hours or 75% of the observed local-hour window).
Paired campus-reference differences
Compare each monitor against the same reference hours.
For every comparator, the difference is calculated hour-by-hour as comparator PM2.5 minus reference PM2.5, using only hours observed at both linked monitors. Positive values mean the comparator monitor was higher during those paired hours.
IHBAS, Dilshad Garden, Delhi - CPCB · 0.7 km from campus
| Comparator | Shared hours | Mean paired difference | Median difference | Comparator higher |
|---|---|---|---|---|
ABVIMS Talkatora Garden, Delhi - DPCC | 137 | +71.1 µg/m³ | +66.0 µg/m³ | 100% |
LHMC Mandir Marg, Delhi - DPCC | 137 | +96.5 µg/m³ | +94.0 µg/m³ | 100% |
AIIMS Sirifort, Delhi - CPCB | 137 | +112.6 µg/m³ | +119.0 µg/m³ | 100% |
Dr BSA Rohini, Delhi - DPCC | 91 | +120.3 µg/m³ | +126.0 µg/m³ | 100% |
VMMC R K Puram, Delhi - DPCC | 119 | +134.7 µg/m³ | +139.0 µg/m³ | 100% |
MAMC ITO, Delhi - CPCB | 10 | +325.3 µg/m³ | +323.5 µg/m³ | 100% |
What this does not mean
This is a paired comparison of nearby ambient reference monitors. It is not an estimate of indoor campus air, personal exposure, a causal effect of attending a campus, or a health ranking. No p-values or confidence intervals are reported while the archive is this young and strongly temporally autocorrelated.
Reproducibility boundary
Live pages and frozen research objects serve different purposes.
This page changes as new hourly observations enter the archive. Formal analyses should cite a defined analysis window and a frozen release or export whenever possible.
Pilot analysis specification 1.2.0
Pilot analyses remain descriptive.
The live pilot analysis specification can evolve during method development. It does not modify frozen protocol v1.1.1. Any primary-period implementation change affecting the frozen estimand, eligibility, QC, uncertainty procedure or interpretation boundary must be documented as a deviation or formal amendment rather than silently absorbed into the pilot spec.