- Practice
- About 22 min
- 5 practice questions
Accuracy, precision and the theory of errors
Distinguish accuracy from precision, classify systematic, random and gross errors, and calculate mean value, absolute error, mean absolute error, relative error and percentage error from repeated readings.
If something looks wrong, please tell us.
By the end of this lesson you can
- Explain accuracy and precision with examples, including the target-shooting picture.
- Classify errors as systematic (five kinds), random or gross, and say how each is reduced.
- Calculate the arithmetic mean of repeated readings and the absolute error of each.
- Find the mean absolute error, relative error and percentage error.
- Write a result in the form a = (aₘ ± Δaₘ) with a unit.
No measurement is perfectly exact. Every reading carries some uncertainty, called an error, and every result calculated from readings inherits it. A good experimenter does not hide errors; she estimates them and reports them with the result.
Accuracy and precision
Accuracy is how close a measurement is to the true value. Precision is how closely repeated measurements agree with one another, or how finely an instrument can resolve a quantity. Suppose a refrigerator's true temperature is 4.0 °C and a thermometer gives 5.1, 5.2, 5.1, 5.0 °C: the readings agree closely (precise) but are about 1 °C off (not accurate). Repeated precise readings are not necessarily accurate, but readings that are consistently accurate are also precise.
Target-shooting picture of accuracy and precision
-
Not accurate, not precise
- Hits scattered widely
- Their average is away from the centre
-
Precise, not accurate
- Hits close together
- The group is away from the centre
-
Accurate and precise
- Hits close together
- The group is at the centre
Resolution matters too. If the true length is about 4.567 cm, a reading of 4.5 cm with a 0.1 cm scale is closer to the truth than 4.63 cm with a 0.01 cm instrument. The first is more accurate but less precise; the second is more precise but less accurate.
Types of error
Systematic errors are reproducible and always push the result the same way (always too high or always too low). They cannot be removed by averaging, and are hard to spot statistically. The textbook lists five kinds:
- Instrumental errors: a badly calibrated instrument or a metre scale with a worn end. Remedy: choose and check instruments carefully.
- Imperfect technique or procedure: for example heat lost by radiation from a poorly insulated calorimeter. Remedy: apply a correction.
- Personal errors: wrong initial setting or careless observation, such as parallax in reading a scale. Remedy: proper precautions.
- External causes: changes in temperature, humidity or pressure during the experiment.
- Least count error: caused by the resolution of the instrument. Remedy: a more precise instrument.
Random (chance) errors come from unpredictable changes in conditions (pressure, temperature, supply voltage) and from small variations in the observer's judgement. They scatter readings on both sides of the true value, so they are reduced by taking many readings and averaging. Gross errors come from sheer carelessness: reading an instrument without setting it properly, recording the wrong value, or using wrong values in calculation. Only an alert, careful observer can avoid them.
Errors in measurement
-
Systematic
Same direction every time
- Instrumental
- Technique or procedure
- Personal
- External causes
- Least count
-
Random
Scatter both ways
- Reduced by averaging many readings
-
Gross
Carelessness
- Avoided only by care and alertness
Error analysis
Let a quantity be measured n times, giving a₁, a₂, ..., aₙ. The arithmetic mean aₘ = (a₁ + a₂ + ... + aₙ)/n is taken as the best estimate of the true value.
- Absolute error of each reading: |Δaᵢ| = |aₘ − aᵢ|.
- Mean absolute error: Δaₘ = (|Δa₁| + |Δa₂| + ... + |Δaₙ|)/n. The true value lies between aₘ − Δaₘ and aₘ + Δaₘ.
- Relative (fractional) error = Δaₘ / aₘ.
- Percentage error = (Δaₘ / aₘ) × 100%.
Result = aₘ ± Δaₘ; relative error = Δaₘ/aₘ; percentage error = (Δaₘ/aₘ) × 100%
Worked example 1: diameter of a wire
Five screw-gauge readings of a wire's diameter (practice data) are 1.42, 1.45, 1.40, 1.44 and 1.39 mm.
| Reading (mm) | Absolute error |1.42 − aᵢ| (mm) |
|---|---|
| 1.42 | 0.00 |
| 1.45 | 0.03 |
| 1.40 | 0.02 |
| 1.44 | 0.02 |
| 1.39 | 0.03 |
Mean: (1.42 + 1.45 + 1.40 + 1.44 + 1.39)/5 = 7.10/5 = 1.42 mm. Mean absolute error: (0.00 + 0.03 + 0.02 + 0.02 + 0.03)/5 = 0.10/5 = 0.02 mm. Relative error: 0.02/1.42 = 0.014. Percentage error: 0.014 × 100% ≈ 1.4%. Result: d = (1.42 ± 0.02) mm.
Worked example 2: a speedometer
A car's speedometer reads 48 km h⁻¹ when the true speed is 50 km h⁻¹. Absolute error = |50 − 48| = 2 km h⁻¹. Relative error = 2/50 = 0.04. Percentage error = 0.04 × 100% = 4%. A percentage error close to zero means the measurement is close to the target value.
Key takeaways
- Accuracy = closeness to the true value; precision = closeness of repeated readings or fineness of resolution.
- Systematic errors are one-directional and are removed by correction or better instruments, not by averaging.
- Random errors are reduced by averaging many readings; gross errors only by care.
- Mean absolute error, relative error and percentage error describe how good a result is.
- Always state a result as (mean ± mean absolute error) with its unit.
Check your understanding
-
1. A thermometer repeatedly reads 37.9 °C, 38.0 °C and 37.9 °C for water whose true temperature is 37.0 °C. The thermometer is:
- accurate and precise
- accurate but not precise
- precise but not accurate
- neither precise nor accurate
Show answer
Answer: C. The readings agree closely with each other (precise) but are about 0.9 °C away from the true value (not accurate). This is typical of a systematic error.
-
2. A metre scale whose end is worn out gives every length too short. This is:
- a random error
- a gross error
- a least count error
- an instrumental (systematic) error
Show answer
Answer: D. The fault is in the instrument and it shifts every reading in the same direction, so it is a systematic error of the instrumental kind. Averaging does not remove it.
-
3. Which error can be reduced by taking a large number of readings and finding their mean?
- Random error
- Instrumental error
- Error due to a worn scale
- Zero error
Show answer
Answer: A. Random errors scatter readings on both sides of the true value, so they partly cancel on averaging. The other three are systematic and push readings one way.
-
4. Readings of a period are 2.0 s, 2.2 s and 2.1 s. The mean absolute error is about:
- 0.1 s
- 0.07 s
- 0.2 s
- 0.03 s
Show answer
Answer: B. Mean = 6.3/3 = 2.1 s. Absolute errors: 0.1, 0.1, 0.0 s. Mean absolute error = 0.2/3 ≈ 0.07 s.
-
5. A length is measured as (25.0 ± 0.5) cm. The percentage error is:
- 0.5%
- 5%
- 2%
- 0.2%
Show answer
Answer: C. Percentage error = (0.5/25.0) × 100% = 0.02 × 100% = 2%.
Sign in to keep track of the lessons you finish.