01
What happened?
Read what is present.
Time series, ranges, basic statistics and visible changes establish what the dataset literally shows.
What does the dataset literally show?
Study / experiment
Studies.
Some questions are easier to understand when you build something and test them.
This is where I experiment with environmental measurements, analysis and interpretation — including the limitations.
Study 01
Reading the Air
Same data. Better questions.
Core question
How much more understanding can better questions extract from the same environmental dataset?
Prominent limitations
What this study cannot claim
- Parts of the dataset were reconstructed/simulated after the original raw data was lost.
- MQ sensor response must not be treated as defensible regulatory gas concentration in ppm.
- Correlation does not establish causation.
- This low-cost setup does not replace calibrated professional environmental instrumentation.
- This is an analysis and interpretation experiment, not regulatory-grade air-quality monitoring.
The setup
Simple hardware.
Interpretation in focus.
Hardware is intentionally simple. The main story is the progression of analysis and interpretation, not the electronics.
- ESP32Microcontroller
- DHT22Temperature and humidity sensor
- MQ-seriesGas-sensor response
- ≈ one weekEnvironmental dataset
Experiment media
The system used in the study.
Five stages
The questions become more demanding.
02
What moves together?
Look for relationships.
Correlation can expose variables that change together and suggest where the next question belongs.
Correlation is a clue, not a cause.
03
What changes the interpretation?
Add context around the number.
Environmental conditions, sensor placement, operating context and missing information can change what a pattern means.
What information is missing from the number itself?
04
Can I trust the interpretation?
Test the evidence boundary.
Sensor limitations, calibration, uncertainty, data quality and reconstruction limitations constrain the analysis.
Better analysis does not repair weak evidence.
- Sensor limitations
- Calibration
- Uncertainty
- Data quality
- Reconstruction limits
05
What could I responsibly decide?
Keep certainty proportional.
The useful endpoint is not the most sophisticated explanation. It is the interpretation or decision the available evidence can actually support.
What does the available evidence actually support?
ObservationPatternHypothesisSupported interpretation / decisionUnsupported conclusion
The sensor did not become more sophisticated. The questions did.
Same data. Better questions.