A threshold is a choice.
Here, three valid readings above 25 °C count as an event only because that is the initial rule. The rule is editable. It is not proof of danger, biological significance or a device fault.
NANO GENETICS / EXPERIMENT LAB
A signal is a beginning.Make it understandable.
Turn a series of measurements into a question you can test. Explore the data, examine its quality and build a record of your next experiment.
This is where our knowledge becomes a working method. Start with a simulated signal or bring a CSV file from a tabletop sensor.
01 / FROM VALUE TO UNDERSTANDING
Your experiment.
Open to inspection.
Change one assumption at a time. The chart, quality review and experiment brief update together. Calculations run in your browser; selected files are not uploaded.
SIGNAL RECORD
See what the data contains.
Static example / 5 simulated readings
Valid readingChosen thresholdGaps remain visible. Missing values are not invented.
- Readings
- 5
- Valid range
- 22.0–22.4 °C
- Valid mean
- 22.2 °C
- Time gaps
- 0
- Flagged readings
- 0
REVIEW RESULT
No sustained threshold event.
In this static example, every value is below 25 °C. The rule describes these readings; it does not explain why they occurred.
Both downloads are branded PDFs with a chart and page numbers. The measurement dossier includes the entire series. Keep your original CSV for reanalysis.
The readings behind the curve.
| time_ms | Value | Quality |
|---|---|---|
| 0 | 22.1 | Valid |
| 1000 | 22.2 | Valid |
| 2000 | 22.0 | Valid |
| 3000 | 22.4 | Valid |
| 4000 | 22.3 | Valid |
CSV format and how data quality is handled
A small, explicit record.
time_ms,value,range_flag,data_kind
0,22.1,within_sensor_range,simulation
1000,22.2,within_sensor_range,simulation
2000,22.0,within_sensor_range,simulationtime_ms must be non-negative and strictly increasing. Values must be finite numbers. Quoted fields, UTF-8 BOM and Windows line endings are accepted. Extra columns are ignored.
Keep quality alongside the value.
The optional range_flag column can contain within_sensor_range or out_of_range. An out-of-range reading is flagged and excluded from the mean, range and threshold rule. The flag comes from the file; this workbench cannot verify a sensor’s specifications.
The optional data_kind column preserves a simulation label. Invalid file structure produces an error without replacing your current dataset. Time gaps and range flags lead to a review result, even if a threshold event is also present.
02 / A METHOD YOU CAN REPEAT
Build a test
someone else can follow.
A useful experiment starts before the first measurement. Make the question, conditions and interpretation explicit, then repeat the process.
- 01
Ask one question.
For example: does the temperature stay above a chosen value after one condition changes? State the quantity, location and observation period before you look at the result.
RECORD: QUESTION + CONDITIONS - 02
Collect with context.
Record the sensor model, units, sampling interval and setup. Keep the original readings. A number without a unit or timestamp is difficult to compare with a later run.
RECORD: RAW DATA + SETUP - 03
Review before interpreting.
Check for gaps and unusable readings first. Then apply the rule you chose. If you change the rule after seeing the result, document that change as a new exploratory choice.
RECORD: QUALITY + CHOSEN RULE - 04
Repeat, then compare.
Repeat the same setup to see how much the result varies. Change one condition and keep a separate record. A reproducible pattern is a starting point for a better explanation.
RECORD: VARIATION + NEXT QUESTION
03 / UNDERSTAND THE RESULT
A pattern deserves
a careful explanation.
We want the connection between a physical signal and a digital conclusion to be visible. That means understanding both what the tool calculates and what the data cannot tell us.
Understand the whole sensor chainConsecutive readings are not a duration.
Three readings describe a count. The time they span depends on their timestamps. This workbench checks a reading sequence and breaks it at a gap; it does not certify how long a physical state lasted.
A tidy curve is not calibrated truth.
A dataset can contain no gaps and still be affected by calibration, placement, noise or drift. This tool checks file structure and the provided flags. It does not independently measure accuracy or uncertainty.
A measurement is not human intent.
A force or acceleration signal tells us about a measured quantity. It does not, by itself, identify a person’s intention or establish a molecular or neurological mechanism. Those questions need their own methods and validation.
Method note: the demo signals and threshold rule are educational examples authored for this workbench. The tool reviews a file or a simulation; it does not connect to devices, make medical assessments or control equipment. Company experiments and product validation are not claimed by the simulated results.
Content and method: Nano Genetics Sweden AB. .