0

Which is the correct relative error?
$$ r_1=\frac{|p_n-p|}{|p_n|} $$ or $$ r_2=\frac{|p-p_n|}{|p|} $$

Dante
  • 1,908

2 Answers2

0

Given some value $u$ and its approximation $u_{approx}$ the definition of elative error is $$\frac{|u-u_{approx}|}{|u|}$$ So if your $p_n$ is the approximation of $p$ in your question the relative error is $r_2$

Ömer
  • 2,038
0

Assuming $p$ is the exact solution and $p_n$ is a numerical approximation, the absolute and relative errors are defined as $$\text{err}_\text{abs} := |p_n-p|$$ and $$\text{err}_\text{rel} := \frac{\text{err}_\text{abs}}{|p|} = \frac{|p_n-p|}{|p|}$$ So $r_2$ is correct.

AlexR
  • 24,905
  • Thank you. But why does Brian Bradie in A Friendly Introduction to Numerical Analysis on the topic of rootfinding methods write the first stopping condition? – Dante Sep 23 '13 at 17:52
  • Basically, as $p_n\to p$, $r_1\to r_2$. It might be easier to compute the error (i.e. faster) or sufficient... – AlexR Sep 23 '13 at 19:00