Concept Architecture
Concept
Theoretically, Floating-Point Arithmetic is the mathematical system used by digital computers to represent and manipulate real numbers over a wide dynamic range. It is founded on finite-precision numerical representation and scientific notation, allowing very large and very small numbers to be stored using a fixed number of binary digits. Floating-point arithmetic forms the computational foundation of numerical analysis, optimisation, simulation and scientific computing.
Mathematically, a floating-point number is represented by a sign, significand (mantissa) and exponent according to the IEEE 754 standard. Arithmetic operations are performed on these finite-precision representations rather than on exact real numbers, introducing rounding errors, loss of significance and other numerical phenomena. Despite these limitations, floating-point arithmetic provides an efficient balance between computational speed, storage requirements and numerical accuracy.
In practice, Floating-Point Arithmetic underlies virtually every computational method used in health economics, including regression analysis, decision-analytic modelling, probabilistic sensitivity analysis, Monte Carlo simulation and optimisation. Understanding floating-point behaviour is essential for ensuring numerical stability, interpreting computational results and designing robust health economic software.
Purpose
Used to represent and manipulate real numbers in digital computation, support numerical algorithms, enable scientific computing and provide the computational foundation for health economic modelling.
Mathematical Formulae
Primary Formula
Floating-Point Representation
x = (?1)? ? m ? ??
where:
- s = sign bit
- m = significand (mantissa)
- ? = numerical base (? = 2 in IEEE 754)
- e = exponent
IEEE Normalised Binary Representation
x = (?1)? ? (1.f) ? 2?
where:
- f = fractional part of the significand
Supporting Formulae
Machine Epsilon
� = smallest positive number such that
1 + � > 1
Floating-Point Relative Error
Relative Error = |x ? fl(x)| � |x|
where:
- x = exact value
- fl(x) = floating-point approximation
Related Mathematical Methods
- Binary representation
- Finite precision
- Machine precision
- Rounding error
- Truncation error
- Numerical stability
Example
The decimal value
0.1
cannot be represented exactly in binary floating-point arithmetic.
Consequently,
0.1 + 0.2
may evaluate internally as
0.30000000000000004
rather than exactly
0.3.
Although the error is extremely small, repeated floating-point operations may accumulate approximation error in large-scale health economic simulations.
Excel Implementation
| Function | Example Formula | Health Economics Application |
|---|---|---|
| Formula | =0.1+0.2 | Demonstrate floating-point representation behaviour. |
| ABS | =ABS((0.1+0.2)-0.3) | Calculate floating-point error. |
| ROUND | =ROUND(A2,12) | Reduce visible floating-point artefacts in reports. |
| EXP | =EXP(A2) | Perform floating-point exponential calculations. |
| LN | =LN(A2) | Perform floating-point logarithmic calculations. |
VBA (Optional)
Automate numerical precision testing, identify floating-point instability, validate computational tolerances and generate numerical diagnostics for health economic models.
Sources
- IEEE. IEEE Standard for Floating-Point Arithmetic (IEEE 754).
- Goldberg D. What Every Computer Scientist Should Know About Floating-Point Arithmetic. ACM Computing Surveys.
- Higham NJ. Accuracy and Stability of Numerical Algorithms.
- Wilkinson JH. Rounding Errors in Algebraic Processes.
- 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 (6)
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 floating-point arithmetic?
The method used by computers to represent and perform calculations on real numbers with finite precision.
Source: Higham NJ. Accuracy and Stability of Numerical Algorithms. 2nd ed. SIAM; 2002. doi:10.1137/1.9780898718027.
What does floating-point arithmetic allow a computer to do?
Floating-point arithmetic allows a computer to represent and perform calculations on real numbers with finite precision, covering a wide range of magnitudes with a fixed number of digits. By scaling numbers with an exponent, it can hold both very large and very small values approximately. This ability to work with real numbers at finite precision is what floating-point arithmetic provides By pairing a fixed number of significant digits with a scale, the scheme spans a wide range of magnitudes, at the cost of holding most real numbers as close approximations rather than exactly.
Source: Higham NJ. Accuracy and Stability of Numerical Algorithms. 2nd ed. SIAM; 2002. doi:10.1137/1.9780898718027.
How does floating-point arithmetic handle real numbers?
Floating-point arithmetic handles real numbers by representing them approximately with finite precision, storing a fixed number of significant digits together with a scale so that many magnitudes can be expressed. Because precision is finite, most real numbers are held as close approximations rather than exactly. This approximate representation is central to how floating-point arithmetic works with real numbers By pairing a fixed number of significant digits with a scale, the scheme spans a wide range of magnitudes, at the cost of holding most real numbers as close approximations rather than exactly.
Source: Higham NJ. Accuracy and Stability of Numerical Algorithms. 2nd ed. SIAM; 2002. doi:10.1137/1.9780898718027.
Why is floating-point arithmetic subject to finite precision?
Floating-point arithmetic is subject to finite precision because it represents real numbers with a fixed number of digits, so values needing more digits are rounded to fit. As the method a computer uses to represent and calculate with real numbers, it inherits the limitation of finite precision. This is why floating-point results carry small approximations rather than exact values By pairing a fixed number of significant digits with a scale, the scheme spans a wide range of magnitudes, at the cost of holding most real numbers as close approximations rather than exactly.
Source: Higham NJ. Accuracy and Stability of Numerical Algorithms. 2nd ed. SIAM; 2002. doi:10.1137/1.9780898718027.
What kinds of values does floating-point arithmetic represent?
Floating-point arithmetic represents real numbers, the values that include fractions and irrational quantities, doing so with finite precision using a fixed number of digits and a scale. This lets a computer approximate a broad range of real values within its arithmetic. Representing real numbers in this way is the purpose of floating-point arithmetic By pairing a fixed number of significant digits with a scale, the scheme spans a wide range of magnitudes, at the cost of holding most real numbers as close approximations rather than exactly.
Source: Higham NJ. Accuracy and Stability of Numerical Algorithms. 2nd ed. SIAM; 2002. doi:10.1137/1.9780898718027.
How does floating-point arithmetic relate to finite precision?
Floating-point arithmetic is the method computers use to represent and calculate with real numbers, while finite precision is the limitation, arising from a fixed number of digits, within which that method operates. Floating-point arithmetic works under finite precision. The two are connected, since finite precision sets the exactness available and floating-point arithmetic is how a computer represents real numbers given it By pairing a fixed number of significant digits with a scale, the scheme spans a wide range of magnitudes, at the cost of holding most real numbers as close approximations rather than exactly.
Source: Higham NJ. Accuracy and Stability of Numerical Algorithms. 2nd ed. SIAM; 2002. doi:10.1137/1.9780898718027.
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-FPA-002
Stable URI · Machine-readable · Resolvable · CC BY 4.0