0:00 / 0:00

Error of Integral Approximations

Approximations aren't perfect - however, we are able to bound the error on them. By increasing nn (the number of rectangles) we can reduce the error created by our area approximations

Error for the Midpoint Approximation

Suppose f(x)<K2|f''\left(x\right)|<K_2 for axba\le x\le b. Then the error using a definite integral Midpoint approximation is

EN=MNActual|E_N|=|M_N-\text{Actual}|
and the error can be bounded by

EN<K2(ba)324n2\boxed{|E_N|<\frac{K_2\left(b-a\right)^3}{24n^2}}


PAGE BREAK

Error for the Trapezoid Approximation

Suppose f(x)<K2|f''\left(x\right)|<K_2for axba\le x\le b. Then the Error using a definite integral Trapezoid approximation is

EN=TNActual|E_N|=|T_N-\text{Actual}|
and the error can be bounded by

EN<K2(ba)312n2\boxed{|E_N|<\frac{K_2\left(b-a\right)^3}{12n^2}}

Error for the Simpson's Approximation

Suppose f(4)(x)<K4|f^{(4)}\left(x\right)|<K_4 for axba\le x\le b. Then the Error using a definite integral Simpson's approximation is
EN=SNActual|E_N|=|S_N-\text{Actual}|
and the error can be bounded by

EN<K4(ba)5180n4\boxed{|E_N|<\frac{K_4\left(b-a\right)^5}{180n^4}}

How large does n have to be to guarantee that the integral approximation of 02x5dx\displaystyle \int_0^2x^5dx
using the Midpoint rule is accurate to within 0.0001?