# The estimation error is strangely obtained from Simpson's 1/3 rule

42 views (last 30 days)
재훈 on 22 May 2024 at 7:23
Commented: 재훈 on 24 May 2024 at 1:49
Hello, I am looking for the estimation error of Simpson's rule of thirds. When dx is 1.01, the integral value is 2908800 and the error is 0.01, but the estimation error is 5.5567e-09. Where did it go wrong?
clc; clear all; close all;
a = -1.93317553181561E-24;
b = 3.788630291530091e-21;
c = -2.3447910280083294e-18
d = -0.019531249999999518;
e = 18.74999999999999
fun = @(t) a.*t.^5 + b.*t.^4 + c.*t.^3 + d.*t.^2+e.*t
x = 0:0.1:960;
fx =fun(x);
n=960;
dx=1.01
int=0;
for i =1:n
plot(x, fx,'k','linewidth',2);
mid=((i-1)+i)/2;
fx_mid = fun(mid);
fx_left = fun(i-1);
fx_right = fun(i);
area_temp = dx/6*(fx_left +4*fx_mid+fx_right);
int = int + area_temp;
x_segment = linspace(i-1, i,100);
Px = fx_left * ((x_segment-mid).*(x_segment-i))/((i-1-mid)*(i-1-i))...
+ fx_mid*((x_segment-i+1)).*(x_segment-i)/((mid-i+1)*(mid-i))...
+ fx_right * ((x_segment-i+1).*(x_segment-mid))/((i-i+1)*(i-mid));
area(x_segment,Px); hold on;
end
C=480;
E_a = -((960.^5)/(2880.*1.01.^4)).*(a.*120.*C+24.*b);%Is there a problem here?
disp('E_a');
disp(E_a);
disp(int);
int_true = 2880000
rel_error=norm(int_true-int)/norm(int_true);
disp('rel_error');
disp(rel_error);
Steven Lord on 23 May 2024 at 13:51

Ayush Anand on 22 May 2024 at 8:11
Edited: Ayush Anand on 22 May 2024 at 8:18
There is a mistake with your formulae for estimation of . The formula for the error estimate in Simpson's rule is generally given by:
where ξ is some number in the interval [a, b], n is the number of intervals, and is the fourth derivative of the function being integrated.
Given your function: ; the fourth derivative of would be:
So, given the interval and with , the formulae for looks like:
Replacing formulae for in your code with this should give you the correct results.
You can check this out for reading more on Simpson's rule: https://www.mathworks.com/matlabcentral/fileexchange/72526-simpson-s-1-3-rule-composite
##### 3 CommentsShow 1 older commentHide 1 older comment
Ayush Anand on 23 May 2024 at 4:49
If the fourth derivative coefficient is very small, the error estimate will indeed be very small. The estimation error provided by Simpson's rule is not always an exact reflection of the true error; it's an estimate that depends on several factors, including the behavior of the function being integrated and the specifics of the numerical method being used. It assumes that the derivative in consideration is relatively constant or at least does not exhibit extreme variations within the interval. For well-behaved functions, where these assumptions hold true, the error estimate can be quite accurate or at least provide a good approximation of the actual error. For functions with significant variations in higher-order derivatives or with singularities, the estimate might not be as reliable.
재훈 on 24 May 2024 at 1:49
Thank you so much for your help. Thanks to this, I learned a little more about estimation error. have a good day! cheers

### Categories

Find more on Numerical Integration and Differential Equations in Help Center and File Exchange

R2024a

### Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!