Accuracy and precision are related but distinct concepts in numerical computation. Accuracy concerns closeness to the true or accepted value. Precision may refer to the resolution of a numerical representation, the number of significant digits or the repeatability of repeated measurements. A result can be highly precise but inaccurate if it consistently differs from the true value. Similarly, a result may be reasonably accurate but represented with limited decimal precision. In numerical computing, the number of stored digits and the accuracy of the computed result depend on both the arithmetic representation and the numerical algorithm.
Accuracy and Precision
Accuracy describes how close a numerical result is to the true value, while precision describes the fineness or repeatability of numerical representation and measurement.
Eₐ = |x − x*|
Explanation
Variables
| Symbol | Meaning |
|---|---|
Eₐ | Absolute error. |
x | True or accepted value. |
x* | Computed or measured value. |
Example or application
If the accepted value is 10, the values 9.99 and 10.01 are both close to the true value. A result such as 12.0000 may be displayed with many digits but is not accurate for this example.
Definition
Accuracy is the closeness of a computed or measured value to the true value. Precision refers to the fineness of numerical representation or the consistency of repeated results, depending on context.
Purpose
To distinguish the concepts of accuracy and precision and understand their effects on numerical results.
Importance
Confusing accuracy with precision can lead to incorrect conclusions about computational quality. A calculation may display many digits without being close to the true value, so both concepts must be understood.
Learning Objectives
Define accuracy and precision.
Explain how the two concepts differ.
Identify precise but inaccurate results.
Understand the role of significant digits.
Relate numerical precision to floating-point representation.
Key Concepts
Accuracy
Precision
Significant digits
Resolution
Repeatability
Absolute error
Numerical representation
Working Principle
Accuracy is evaluated by comparing a result with a known or accepted value. Precision is assessed through the numerical resolution or consistency of the representation and repeated calculations.
Process
Obtain a numerical result.
Determine the accepted value if available.
Calculate the error to assess accuracy.
Inspect the number of represented significant digits.
If repeated measurements exist, examine their spread.
Report accuracy and precision appropriately.
Applications
Scientific measurements
Laboratory experiments
Floating-point computation
Engineering tolerances
Numerical algorithm evaluation
Data acquisition
Worked Example
Problem
Compare two approximations of the true value 5.000.
Steps
Approximation A = 4.99.
Approximation B = 5.20.
Absolute error of A = |5.000 − 4.99| = 0.01.
Absolute error of B = |5.000 − 5.20| = 0.20.
Approximation A is more accurate because it has the smaller absolute error.
Answer
Approximation A is more accurate.
Common Misconceptions
More decimal places do not automatically mean greater accuracy.
Accuracy and precision are not interchangeable terms.
A result can be precise but inaccurate.
The meaning of precision can vary between measurement science and numerical representation.
Practice Questions
| Question | Answer |
|---|---|
What is accuracy? | The closeness of a result to the true or accepted value. |
Can a result be precise but inaccurate? | Yes. A result may be consistently represented or repeated while remaining far from the true value. |
Which is more accurate for a true value of 20: 19.9 or 20.5? | 19.9, because its absolute error is 0.1 rather than 0.5. |
Key Takeaway
Accuracy measures closeness to the truth, while precision concerns numerical detail or repeatability; a reliable computation requires attention to both.
Prerequisites
SCI-MAT-014-011
Basic decimal arithmetic
Significant digits
Reference
Higham, N. J. Accuracy and Stability of Numerical Algorithms.
NIST/SEMATECH e-Handbook of Statistical Methods.
Related Topics
SCI-MAT-014-011
SCI-MAT-014-016
SCI-MAT-014-019
SCI-MAT-014-020
Sources and verification
Sources are shown as provided in the project data. Not every source has yet been independently verified. Please report errors or missing citations.
- Source verification pending; check an authoritative textbook, university, standards body, or peer-reviewed source.