The Trapezoid method for [[Quadrature]] is a numerical integration method that uses the [[Left-Endpoint Method]] and [[Right-Endpoint Method]] together. $\large \begin{align} A = h\pa{ \frac{1}{2}f(a) + f(a+h) +f(a+2h) + \cdots + f(b-h)+ \frac{1}{2}f(b) } \end{align}$ This method uses the average of the left endpoint and right endpoint method. $\huge T = \frac{L_{}+R}{2} $ The logic behind this method is if $f'$ is positive then the [[Left-Endpoint Method]] will be too low, but the [[Right-Endpoint Method]] will be too high. ### [[Numerical Analysis Error Bounds|Error]] Lets say we can model our [[Function]] as some quadratic function. Let $f(x)=k_{1}x+k_{2}x^{2}$ with some step side $h>0$. $\huge \begin{align} L&= b -a\\ n &= \frac{L}{h} \end{align} $ $\huge \begin{align} f(0) &= 0\\ f(h) &= k_{1}h + k_{2}h^{2}\\ A_{T}&=\frac{1}{2}\pa{f(0)+f(h)} h \\ &= \frac{k_{1}h^{2}}{2} + \frac{k_{2}h_{3}}{2} \end{align}$ The actual integral would equal: $\huge \int_{0}^{h}f(x)\d x = \frac{k_{1}h^{2}}{2}+ \frac{k_{2}h_{3}}{3} $ Our error (for one step) would be the difference between these, $\huge \begin{align} \epsilon_{1} &= \pa{\frac{k_{1}h_{2}}{2}+\frac{k_{2}h_{3}}{3}} - \pa{\frac{k_{1}h_{2}}{2}+\frac{k_{2}h_{3}}{2}}\\ &= \frac{k_{2}h_{3}}{6} \end{align}$ Or, $\huge \epsilon = \frac{h^{3}}{12} f'' $ For the general case, $\huge \epsilon_{1} \leq \frac{h^{3}}{12}|f''_{\op{ext}}| $ The total error bounded would be: $\huge \epsilon \leq n\frac{h^{3}}{12}|f_{\op{ext}}'' = \frac{h^{2}L}{12}|f''_{\op{ext}}| $