Numerical computations may contain several types of error. Inherent data errors arise from uncertainty or inaccuracies in the input. Truncation errors occur when an infinite process or mathematical expression is replaced by a finite approximation. Rounding errors arise because computers store numbers with finite precision. Discretization errors occur when continuous problems are represented on finite grids. These errors may interact and propagate through later calculations. Error analysis identifies their sources, estimates their magnitudes and studies how they influence the final result. It is an essential part of designing reliable numerical algorithms.
Error Analysis in Numerical Computation
Error analysis studies the sources, propagation and effects of errors in numerical calculations to determine the reliability of computed results.
Eₜ = x − x*
Explanation
Variables
| Symbol | Meaning |
|---|---|
Eₜ | Signed total error under the stated comparison. |
x | Exact or true value. |
x* | Computed approximate value. |
Example or application
Approximating sin(x) using a finite Taylor polynomial introduces truncation error because the infinite series is replaced by a finite number of terms.
Definition
Error analysis is the mathematical study of the sources, magnitudes and propagation of errors introduced during numerical approximation and computation.
Purpose
To identify and quantify numerical errors and understand their influence on the reliability of computational results.
Importance
Error analysis is necessary because numerical calculations may amplify small inaccuracies. It helps researchers select appropriate algorithms, precision levels and stopping criteria, and determine whether a result is sufficiently reliable.
Learning Objectives
Define numerical error analysis.
Identify major sources of numerical error.
Distinguish truncation, rounding and input errors.
Understand error propagation.
Explain how error estimates support reliable numerical computation.
Key Concepts
Error analysis
Truncation error
Rounding error
Input error
Discretization error
Error propagation
Error bound
Working Principle
Error analysis compares an approximate computation with a known value or a theoretical error bound. It identifies the contributions of different error sources and studies how they influence subsequent operations.
Process
Identify the mathematical quantity being calculated.
Determine the sources of possible error.
Estimate input and representation errors.
Analyze approximation and discretization errors.
Study how errors propagate through operations.
Combine suitable error bounds when mathematically justified.
Assess whether the final error meets the required tolerance.
Applications
Numerical integration
Differential equation solvers
Scientific simulations
Floating-point software
Engineering calculations
Numerical optimization
Worked Example
Problem
Approximate eˣ at x = 1 using the first three terms of its Taylor series and identify the truncation error.
Steps
The Taylor series is eˣ = 1 + x + x²/2! + x³/3! + ...
Using the first three terms gives 1 + 1 + 1/2 = 2.5.
The exact value is e ≈ 2.718281828.
The absolute difference is approximately |2.718281828 − 2.5| = 0.218281828.
This difference is the approximation error for the stated three-term calculation.
Answer
The three-term approximation is 2.5, with an absolute error of approximately 0.218281828.
Common Misconceptions
Not all numerical errors originate from rounding; approximation and input errors also matter.
Truncation error and rounding error are different sources of error.
Small individual errors may accumulate or become amplified.
A numerical method should be analyzed under the assumptions and conditions for which its error estimates are valid.
Practice Questions
| Question | Answer |
|---|---|
What is truncation error? | Error introduced when an infinite process, series or mathematical procedure is replaced by a finite approximation. |
What causes rounding error? | The finite precision used to represent and calculate numerical values. |
Why is error propagation important? | Because errors introduced in early calculations may influence later operations and the final result. |
Key Takeaway
Error analysis identifies where numerical inaccuracies originate, how they propagate and whether the final computed result is reliable.
Prerequisites
SCI-MAT-014-011
SCI-MAT-014-012
SCI-MAT-014-016
SCI-MAT-014-023
Reference
Higham, N. J. Accuracy and Stability of Numerical Algorithms.
Burden, R. L., Faires, J. D., and Burden, A. M. Numerical Analysis.
Related Topics
SCI-MAT-014-011
SCI-MAT-014-013
SCI-MAT-014-023
SCI-MAT-014-026
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.