HEAT-SAFE HOMES — DATA AND METHODS

The CSV uses one row per home and timestamp:
site_id,home_id,intervention,timestamp,indoor_c,outdoor_c,humidity_pct,occupied

Use non-identifying home and site IDs. The intervention label should stay the
same for each home in a file. Use ISO timestamps with a UTC offset, such as
2026-07-15T20:00:00-07:00 or 2026-07-16T03:00:00Z. Use Celsius for temperatures,
percentage for relative humidity, and 1 or 0 for occupied; blank means unknown.
An identical instant entered twice for the same home is an error, even when the
time-zone offsets differ. Blank measurements are missing, never zero.

The analysis uses a user-defined window, reference temperature, fixed UTC offset
and maximum gap between adjacent readings. It holds each earlier measurement
constant until the next eligible measurement and excludes intervals with missing
endpoints or gaps longer than the selected maximum. It does not extrapolate
before the first or after the final reading. Hours above the reference and
degree-hours cover only observed intervals. The reference temperature is an
arbitrary analytical parameter, not a health or safety threshold. Coverage and
missing counts appear with the results.

Night is 22:00–06:00 at the selected fixed offset. Analyze different offsets in
separate windows around daylight-saving transitions. An interval is classified
as occupied only when both endpoint flags are 1. Relative humidity is time
weighted on valid intervals and its coverage is independent of temperature.
Relative humidity alone cannot establish mold risk, moisture damage or safety.

Two homes must be at the same site to compare them. The tool compares only
identical original sensor interval endpoints where both indoor and outdoor
readings are present and the outdoor readings differ by no more than 0.5 °C at
either endpoint. It shows matching hours and their share of the requested window.
Timing and weather matching do not account for differences in the buildings,
occupant behavior or other conditions. A descriptive difference does not prove
that an intervention caused it. Repeated readings are not independent homes.

SELECTED PRIMARY SOURCES

Vellingiri and colleagues (2020), sixteen Ahmedabad households:
https://www.mahilahousingtrust.org/wp-content/uploads/2022/06/MHT_Combating-Climate-Change%E2%80%90induced-Heat-Stress.pdf
Different roof types were compared in short daytime measurements. Homes and
weather differ; the study does not establish a universal retrofit effect.

Rosado and colleagues (2014), Lawrence Berkeley National Laboratory, two Fresno
show homes:
https://eta-publications.lbl.gov/sites/default/files/measured_temperature_reductions_and_every_savings_from_a_cool_tile_roof_on_a_central_california_home.pdf
Roof material, ventilation and thermal mass differed. An attic-temperature
result should not be represented as an occupied-room temperature result.

Sharmin and Rahman, 2023 conference proceedings, one Ahmedabad home:
https://carbse.org/CATE2023/PaperID_1166_Optimising_energy_efficiency_and_thermal_comfort_measures_for_a_low-income_residential_building_in_Ahmedabad%2C_India.pdf
The three-month ventilation comparison was simulated using a model informed by
short field measurements; it was not a field-measured intervention effect.

This tool processes chosen input in your browser. No field pilot or safety
assessment is established by the included synthetic example.