Correct the Instrument Before the Claim

Correct the Instrument Before the Claim

Ada re-checks the co-located offsets, the grid coverage, and the commuter-street concentration before the aggregate is allowed to claim anything

foundations
math-foundations
participatory-sensing
calibration
beginner
Ada ADA · CALCULATION AUDIT

Correct the Instrument Before the Claim

A crowd-sourced noise map is only as trustworthy as the phones that feed it, and every phone is a different, uncalibrated instrument. Placed beside a reference meter reading 64 dB, family A reads 65 dB and family B reads 68 dB — offsets of +1 and +4 dB — so a later B reading of 72 dB should be corrected to 68 dB before it joins the aggregate. This audit works that co-located calibration step to ask why you must correct the instrument before trusting the claim.

Companion to the chapter Participatory Sensing — every number here comes from that chapter.

See the relationship before changing it

The figure reads from left to right. The blue card is raw family b reading. The middle card applies the page rule. The green card is corrected reading. Walk the arrows once: set the input, apply the rule, then read the result with its unit.

Raw family B reading changes corrected reading An input card leads through the rule corrected = raw reading - 4 dB family offset to the corrected reading result. INPUT PAGE INPUT APPLY THE RULE predict calculate check units OUTPUT RESULT
Walk the arrows. Co-location removes the handset offset before readings enter the shared map.

Derive the baseline in four named moves

  1. 1

    Name the input. The chapter baseline is 72 dB.

  2. 2

    Name the relationship. corrected = raw reading - 4 dB family offset

  3. 3

    Substitute with units. 72 - 4 = 68.00 dB

  4. 4

    Read the result. Keep the unit beside the value. Use it only inside the technical boundary on this page.

Predict, then change raw family b reading

Try Predict the direction of corrected = raw reading - 4 dB family offset. Test another raw family b reading, then compare corrected reading.

72 dB
Chapter baseline
Corrected reading

Observe Co-location removes the handset offset before readings enter the shared map. Reset raw family b reading to 72 and compare corrected reading.

Explain Co-location removes the handset offset before readings enter the shared map.

Check yourself

What should you do before trusting a moved-control result?
Answer: Predict its direction, apply the shown relationship, keep the units, and reset to the worked baseline.
What does this small model leave out?
Answer: Only raw family b reading moves here. Field effects named in the technical boundary stay fixed.

Honesty note. This teaching model changes only raw family b reading. It is not a field prediction.

TryAt Check derivation, compare phone families A and B with the 64 dB reference, then load B's later 72 dB reading.
ObserveFamily offsets resolve to +1 and +4 dB, so subtracting B's offset changes its raw 72 dB observation to 68 dB.
ExplainCo-location estimates each instrument's additive bias; the +4 dB correction prevents handset calibration from masquerading as environmental noise.

Ready: use the stated baseline inputs, then compare each displayed result.

Check the offsets, the coverage, and the concentration

  • Family A offset: 65 dB - 64 dB = +1 dB, so A is about 1 dB high in this check.
  • Family B offset: 68 dB - 64 dB = +4 dB, so B is about 4 dB high in this check.
  • Later B adjusted value: 72 dB - 4 dB = 68 dB. The raw 72 dB remains stored, but the aggregate uses the corrected 68 dB value.
  • Coverage touched: 4 cells / 12 cells = 0.333..., about 33%. Repeated windows: 2 cells / 12 cells = 0.166..., about 17%.
  • Commuter-street concentration: 150 records / 180 records = 0.833..., about 83% of the accepted records from two streets in one hour.

Those checks explain the physics and the statistics at the same time: decibel readings need instrument correction before comparison, and a large sample count still cannot support a citywide claim when coverage is only about one third of the grid and repeated evidence covers about one sixth.

Every number above is taken from the chapter’s own worked example and re-derived step by step.

Technical boundaries. Participant selection, spatial bias, sensor ageing, nonlinear microphone response, and missing uploads are omitted; the fixture applies one additive co-location offset to fixed dB readings.