Concept Architecture
Concept
Theoretically, Numerical Stability is the property of a numerical algorithm whereby small perturbations arising from rounding errors, measurement errors or finite-precision arithmetic do not grow excessively during computation. It is a fundamental concept in numerical analysis because it determines whether an algorithm produces reliable approximations to the true mathematical solution. Numerical stability concerns the behaviour of the computational method itself, distinguishing it from conditioning, which describes the sensitivity of the underlying mathematical problem.
Mathematically, a numerically stable algorithm controls the propagation and amplification of computational errors throughout successive calculations. Stable algorithms produce solutions whose errors remain comparable to the unavoidable errors introduced by finite-precision arithmetic, whereas unstable algorithms amplify these errors, leading to inaccurate or divergent results. Backward stability and forward stability are commonly used theoretical frameworks for evaluating algorithmic performance.
In practice, Numerical Stability is essential in health economics because regression estimation, optimisation routines, decision-analytic models, Markov simulations and probabilistic sensitivity analyses all involve extensive numerical computation. Stable algorithms improve the reliability of estimated costs, utilities, survival probabilities and incremental cost-effectiveness ratios, ensuring that computational artefacts do not influence healthcare decision-making.
Purpose
Used to evaluate the reliability of numerical algorithms, minimise computational error, support accurate model implementation and ensure robust health economic analyses.
Mathematical Formulae
Primary Formula
Forward Relative Error
Forward Error = ?x? ? x? � ?x?
where:
- x = exact solution
- x? = computed solution
Supporting Formulae
Backward Relative Error
Backward Error = ?�A? � ?A?
Conditioning Relationship
Forward Error � ?(A) ? Backward Error
where:
- ?(A) = matrix condition number
Related Mathematical Methods
- Conditioning
- Ill-conditioned matrix
- Floating-point arithmetic
- Error propagation
- Rounding error
- Backward error analysis
Example
Suppose two numerical algorithms solve the same system of equations.
Algorithm A returns a solution differing from the exact solution by
0.000001
after thousands of arithmetic operations.
Algorithm B produces an error of
0.25
for the same problem because rounding errors accumulate throughout the computation.
Algorithm A is therefore numerically stable, whereas Algorithm B is numerically unstable.
Excel Implementation
| Function | Example Formula | Health Economics Application |
|---|---|---|
| ABS | =ABS(B2-A2) | Calculate numerical solution error. |
| Formula | =ABS((B2-A2)/A2) | Calculate forward relative error. |
| MINVERSE | =MINVERSE(A2:C4) | Evaluate matrix computations for numerical stability. |
| MMULT | =MMULT(A2:C4,D2:F4) | Perform stable matrix calculations. |
| IF | =IF(C2<1E-8,"Numerically Stable","Review Algorithm") | Assess computational stability against a predefined tolerance. |
VBA (Optional)
Automate monitoring of numerical error growth, evaluate algorithm stability, detect unstable computations and generate diagnostics supporting reliable implementation of health economic simulation and optimisation models.
Sources
- Higham NJ. Accuracy and Stability of Numerical Algorithms.
- Wilkinson JH. Rounding Errors in Algebraic Processes.
- Trefethen LN, Bau D. Numerical Linear Algebra.
- Golub GH, Van Loan CF. Matrix Computations.
- Press WH, Teukolsky SA, Vetterling WT, Flannery BP. Numerical Recipes: The Art of Scientific Computing.
- Briggs A, Claxton K, Sculpher M. Decision Modelling for Health Economic Evaluation.
Related Concepts (5)
Library
Publications
1
Accuracy and Stability of Numerical Algorithms — Nicholas J. Higham, 2nd Edition ed., 2002 (Society for Industrial and Applied Mathematics (SIAM))
The standard reference on floating-point error, conditioning, stability and the numerical reliability of matrix, polynomial and linear-algebra algorithms.
BookView source →
Frequently Asked Questions (6)
What is numerical stability?
The property of a numerical method that limits the growth and amplification of rounding and computational errors during calculations.
Source: NIST Digital Library of Mathematical Functions. Chapter 3: Numerical Methods.
What does numerical stability describe about a method?
Numerical stability describes a property of a numerical method: that it limits the growth and amplification of rounding and computational errors during calculations. A stable method keeps small errors small as the computation proceeds, rather than letting them swell. This control over the growth of errors is what numerical stability captures about a method Because a stable method keeps small errors from swelling as a computation runs, it protects the accuracy of the result, which is why stability is weighed alongside speed when choosing a method.
Source: NIST Digital Library of Mathematical Functions. Chapter 3: Numerical Methods.
What errors does numerical stability keep in check?
Numerical stability keeps rounding and computational errors in check, limiting their growth and amplification during calculations. These are the small errors that arise from finite-precision arithmetic and the steps of the computation. Preventing such errors from being amplified is the concern of numerical stability Because a stable method keeps small errors from swelling as a computation runs, it protects the accuracy of the result, which is why stability is weighed alongside speed when choosing a method.
Source: NIST Digital Library of Mathematical Functions. Chapter 3: Numerical Methods.
Why is numerical stability desirable?
Numerical stability is desirable because it limits the growth and amplification of rounding and computational errors, so a stable method yields results whose accuracy is not destroyed by the accumulation of small errors. Without it, minor errors could swell and dominate the outcome. This protection of accuracy is why numerical stability is sought in a method Because a stable method keeps small errors from swelling as a computation runs, it protects the accuracy of the result, which is why stability is weighed alongside speed when choosing a method.
Source: NIST Digital Library of Mathematical Functions. Chapter 3: Numerical Methods.
Is numerical stability a property of the method or the problem?
Numerical stability is a property of the numerical method, since it concerns how the method limits the growth and amplification of rounding and computational errors during calculations. A given problem may be solved by a stable or an unstable method. This attachment to the method rather than the problem is central to what numerical stability means Because a stable method keeps small errors from swelling as a computation runs, it protects the accuracy of the result, which is why stability is weighed alongside speed when choosing a method.
Source: NIST Digital Library of Mathematical Functions. Chapter 3: Numerical Methods.
How does numerical stability relate to rounding error?
Numerical stability is the property of a method that limits the growth and amplification of rounding and computational errors, while rounding error is the difference introduced when a value is rounded to fit a limited number of digits. Rounding error is one of the errors a stable method keeps from swelling. The two are connected, since numerical stability governs how rounding errors behave through a calculation Because a stable method keeps small errors from swelling as a computation runs, it protects the accuracy of the result, which is why stability is weighed alongside speed when choosing a method.
Source: NIST Digital Library of Mathematical Functions. Chapter 3: Numerical Methods.
Trust Record
Verified by Dr Darrin Baines
British health economist
Professional identity: darrinbaines.org
Verification date: 26 May 2026
Content version: 1.0.0
Canonical Identity
- Persistent URI
- https://healtheconomics.wiki/concept/numerical-stability
- Term code
- CS-NA-ST-002
Stable URI · Machine-readable · Resolvable · CC BY 4.0