Concept Architecture
Concept
Theoretically, an Ill-Conditioned Matrix is a matrix for which small perturbations in its elements can produce disproportionately large changes in the solution of associated linear systems. It is a fundamental concept in numerical linear algebra and numerical analysis because it characterises the sensitivity of matrix computations to measurement error, rounding error and finite-precision arithmetic. Ill-conditioning is a property of the mathematical problem itself rather than the algorithm used to solve it.
Mathematically, the degree of ill-conditioning is quantified by the matrix condition number, which measures the amplification of relative input errors into relative solution errors. A matrix with a condition number close to one is considered well-conditioned, whereas a matrix with a very large condition number is ill-conditioned. Ill-conditioned matrices frequently arise when columns are nearly linearly dependent or when the determinant approaches zero.
In practice, Ill-Conditioned Matrices occur in health economics during regression analysis, parameter estimation, Markov modelling, optimisation, calibration and simulation. Detecting ill-conditioning is essential because numerical solutions may become unstable, leading to unreliable parameter estimates, poor model convergence and inaccurate health economic predictions.
Purpose
Used to identify numerically unstable matrix problems, assess sensitivity of linear systems, evaluate computational reliability and support robust implementation of health economic models.
Mathematical Formulae
Primary Formula
Matrix Condition Number
?(A) = ?A? ? ?A???
where:
- A = coefficient matrix
- ?�? = matrix norm
Supporting Formulae
Relative Error Bound
Relative Solution Error � ?(A) ? Relative Input Error
Determinant
det(A) � 0
often indicates potential ill-conditioning, although it is not a definitive criterion.
Related Mathematical Methods
- Conditioning
- Condition number
- Matrix inversion
- Linear algebra
- Numerical stability
- Singular Value Decomposition
Example
Suppose a regression model produces a coefficient matrix with
?(A) = 25,000
A measurement error of only
0.01%
may therefore produce an approximate solution error of
25,000 ? 0.0001
= 2.5
or 250%.
The matrix is therefore ill-conditioned, indicating that the computed parameter estimates may be numerically unreliable despite only minor errors in the original data.
Excel Implementation
| Function | Example Formula | Health Economics Application |
|---|---|---|
| MINVERSE | =MINVERSE(A2:C4) | Calculate the inverse of a coefficient matrix. |
| MMULT | =MMULT(A2:C4,D2:F4) | Perform matrix multiplication during numerical analysis. |
| MDETERM | =MDETERM(A2:C4) | Identify matrices approaching singularity. |
| ABS | =ABS((B2-A2)/A2) | Calculate relative error amplification. |
| Solver | Evaluate numerical stability of optimisation models. | Assess matrix conditioning in calibration and regression analyses. |
VBA (Optional)
Automate estimation of matrix condition numbers, detect ill-conditioned matrices, identify unstable numerical problems and generate diagnostics supporting robust health economic modelling.
Sources
- Higham NJ. Accuracy and Stability of Numerical Algorithms.
- Golub GH, Van Loan CF. Matrix Computations.
- Trefethen LN, Bau D. Numerical Linear Algebra.
- Burden RL, Faires JD. Numerical Analysis.
- 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 (4)
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 an ill-conditioned matrix?
A matrix for which small changes in its elements or input values can produce disproportionately large changes in the computed solution.
Source: NIST Digital Library of Mathematical Functions. Chapter 3: Numerical Methods.
What characterises an ill-conditioned matrix?
An ill-conditioned matrix is characterised by high sensitivity, since small changes in its elements or input values can produce disproportionately large changes in the computed solution. Problems involving such a matrix are difficult to solve accurately, because tiny disturbances are amplified. This amplification of small changes into large ones is what marks a matrix as ill-conditioned Because the sensitivity lies in the matrix itself, no algorithm can fully rescue accuracy, so recognising an ill-conditioned matrix helps explain why a computed solution may be unreliable.
Source: NIST Digital Library of Mathematical Functions. Chapter 3: Numerical Methods.
Why is an ill-conditioned matrix difficult to work with?
An ill-conditioned matrix is difficult to work with because small changes in its elements or input values can produce disproportionately large changes in the computed solution, so even slight rounding or measurement error can badly distort the result. Accurate answers become hard to obtain regardless of the method used. This vulnerability to small disturbances is why an ill-conditioned matrix poses a problem Because the sensitivity lies in the matrix itself, no algorithm can fully rescue accuracy, so recognising an ill-conditioned matrix helps explain why a computed solution may be unreliable.
Source: NIST Digital Library of Mathematical Functions. Chapter 3: Numerical Methods.
What happens to the solution when an ill-conditioned matrix is perturbed?
When an ill-conditioned matrix is perturbed, the computed solution can change disproportionately, since small changes in its elements or input values produce large changes in the result. A minor adjustment to the inputs may shift the answer far more than expected. This outsized response of the solution to small perturbations is the defining behaviour of an ill-conditioned matrix Because the sensitivity lies in the matrix itself, no algorithm can fully rescue accuracy, so recognising an ill-conditioned matrix helps explain why a computed solution may be unreliable.
Source: NIST Digital Library of Mathematical Functions. Chapter 3: Numerical Methods.
Does the difficulty of an ill-conditioned matrix depend on the method used?
The difficulty of an ill-conditioned matrix does not depend on the method used, because the sensitivity lies in the matrix itself, where small changes in its elements or input values produce disproportionately large changes in the solution. No algorithm can fully overcome this inherent sensitivity. The difficulty is therefore a property of the ill-conditioned matrix rather than of any particular solution method Because the sensitivity lies in the matrix itself, no algorithm can fully rescue accuracy, so recognising an ill-conditioned matrix helps explain why a computed solution may be unreliable.
Source: NIST Digital Library of Mathematical Functions. Chapter 3: Numerical Methods.
How does an ill-conditioned matrix relate to conditioning?
An ill-conditioned matrix is one for which small changes in its elements or input values produce disproportionately large changes in the computed solution, while conditioning is the sensitivity of a mathematical problem to such small changes. An ill-conditioned matrix is the case of poor conditioning applied to a matrix. The two are connected, since conditioning is the general property and an ill-conditioned matrix is a matrix exhibiting a poor form of it Because the sensitivity lies in the matrix itself, no algorithm can fully rescue accuracy, so recognising an ill-conditioned matrix helps explain why a computed solution may be unreliable.
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: 2 Apr 2026
Content version: 1.0.0
Canonical Identity
- Term code
- CS-NA-ST-004
Stable URI · Machine-readable · Resolvable · CC BY 4.0