Canonical: https://knowlpedia.com/asymptotics/remainder-estimate/

# Remainder estimate

A quantitative error bound after subtracting a stated approximation.

For an approximation \(S_\varepsilon\) to \(F_\varepsilon\), a **remainder estimate** bounds \(R_\varepsilon=F_\varepsilon-S_\varepsilon\), for example
\[
\|R_\varepsilon\|_X\le C\varepsilon^a.
\]
It must identify the [norm](https://knowlpedia.com/linear-algebra/normed-vector-space/) or seminorm, the parameter range, exponent, and constant dependencies. In an expansion, \(S_\varepsilon\) can be a [finite truncation](https://knowlpedia.com/asymptotics/finite-truncation/).

## Higher order and flat errors

Relative to a nonzero term of order \(\varepsilon^b\), an error bound with \(a>b\) is higher order. An error is flat in the parameter if, for every \(N\), it is \(O(\varepsilon^N)\) in the specified topology. A single positive-order bound does not imply flatness, nor does a value bound imply estimates for derivatives.

## Derivative topology

[Infinite-order decay with derivatives](https://knowlpedia.com/asymptotics/infinite-order-decay-with-derivatives/) makes the all-orders meaning precise for a positive scale function.
