Why does making a finite-difference step extremely small sometimes worsen a derivative estimate?
MathematicsLevel: IntermediateDerivative and probability numerics
Why does making a finite-difference step extremely small sometimes worsen a derivative estimate?
Hints
- Consider f(x+h)−f(x).
- Floating-point precision is finite.
Answer
Subtracting nearly equal function values amplifies rounding error. Step size balances truncation error against cancellation; smaller is not always more accurate.
Source and licence: SciPy — numerical differentiation · BSD-3-Clause · answer: Quantoros editorial