As far as I can tell, if you start with the integral form of the remainder of a Taylor polynomial then you can derive the Lagrange form by an application of the mean value theorem for integrals. From Spivak's Calculus:

If $f^{\left({n + 1}\right)}$ is continuous on $[a, x]$, then

$$\displaystyle R_{n,a} \left({x}\right) = \int_a^x \dfrac {f^{\left({n + 1}\right)}\left({t}\right)}{n!} \left({x - t}\right)^n \, \mathrm d t$$

Let m and M be the minimum and maximum of $\dfrac {f^{\left({n + 1}\right)}}{n!}$ on $[a, x]$, then $R_{n,a}(x)$ satisfies

$$m\int_a^x \left({x - t}\right)^n \, \mathrm d t \le R_{n,a} \left({x}\right) \le M\int_a^x \left({x - t}\right)^n \, \mathrm d t$$

so we can write

$$\displaystyle R_{n,a} \left({x}\right) = \alpha \cdot \dfrac {\left({x - a}\right)^{n+1}} {{n+1}!}$$

Now here is my question. Shouldn't this imply that there is a number $x^* \in \left[{a \,.\,.\, x}\right]$ such that

$$\displaystyle R_{n,a} \left({x}\right) = \dfrac {f^{\left({n + 1}\right)} \left({x^*}\right)} {\left({n + 1}\right)!} \left({x - a}\right)^{n + 1}$$

in accordance with the mean value theorem for integrals?

Yet all sources I reviewed, including Spivak, state that $x^* \in \left({a \,.\,.\, x}\right)$.

  • 1
    $\begingroup$ For sure it is not wrong to consider the closed interval... It would be nice though to come up with an example where $x^*=a$. $\endgroup$ – PierreCarre Feb 4 at 16:26

If for instance $f(x)$ is a 2nd degree polynomial, the 1st order Taylor approximation around $x=0$ tells you that $$ f(x)=f(0) + f'(0)x + \frac{f''(x^*)}{2!} x^2. $$

But of course, since the second order approximation is exact, we have that $x^*=0$. So, for sure, it can happen that $x^*$ is in boundary.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.