Guide to Error Propagation
1 Why Quote Errors?
- Every measurement has a limit (
±
value) that brackets where the actual value can lie:contentReference[oaicite:1]{index=1}. - Instrument error is set by device design (e.g.\ ruler ±0.5 mm, micrometer ±0.001 mm) and user skill:contentReference[oaicite:3]{index=3}.
2 Types of Error
Type | Definition | Typical Size |
---|---|---|
Instrument (reading) | Finite resolution of the device | Specified by manufacturer |
Statistical | Scatter of repeated readings (includes instrument + natural variation) | Use sample standard deviation |
These notes follow the worst-case (maximum) method; the same algebra works with 1 σ uncertainties if you treat σ like Δx:contentReference[oaicite:5]{index=5}.
3 Expressing Error
- Absolute error: (same units).
- Relative error: (dimensionless).
- Percentage error: :contentReference[oaicite:7]{index=7}.
Example – ruler reading: gives an 8 % relative error.
4 Propagation Rules
Rule 1 Addition & Subtraction
If with and then
Example: from mm +
Subtractions use the same formula; relative error often grows (4.5 mm ± 0.6 mm → 13 %):contentReference[oaicite:11]{index=11}.
Rule 2 Multiplication & Division
For or add relative errors:
Example volume: → 17 % error from adding three relative
Rule 3 General Function
For small
With m and ,
:contentReference[oaicite:17]{index=17}.
For multivariable sum the partial-derivative contributions:contentReference[oaicite:19]{index=19}.
Special Cases
- Compare two results: subtract and add errors (Rule 1):contentReference[oaicite:21]{index=21}.
- Multiply/divide by an exact constant: scale the absolute error by that constant:contentReference[oaicite:23]{index=23}.
- Trig/log functions: convert angles to radians before using Rule 3:contentReference[oaicite:25]{index=25}.
5 Confidence Levels (Assuming Gaussian Errors)
- About 68 % of readings lie within ,
- 95 % within ,
- 99.7 % within :contentReference[oaicite:27]{index=27}.
6 Reporting Guidelines
- Quote one, at most two, significant figures in the error:contentReference[oaicite:29]{index=29}.
- Round the measured value to the same decimal place as its error.
- Use % errors when specifications are given that way (common in electronics):contentReference[oaicite:31]{index=31}.
Quick Reference Sheet
Operation | Propagation Formula | When to Use |
---|---|---|
Adding lengths, masses, voltages | ||
or | Areas, volumes, ratios | |
$\Delta G \approx | f’(X) | |
$\Delta G \approx | f_X |
Tip: Always check whether worst-case propagation is appropriate; for large datasets, root-sum-of-squares (RSS) with σ may be more realistic.