I'm tasked with determining the values of $x$ for which $f(x)=\cos{x}\approx 1-\frac{x^2}{2!}+\frac{x^4}{4!}$ has an error no greater than 0.001.

Using the error for Taylor polynomials $E = |R_n(x)| = \frac{(x-c)^{n+1}}{(n+1)!}\max|f^{n+1}(x)|$

I've determined that $$E = \frac{x^5}{5!}\le0.001$$

since $f^{(5)}(z)=-\sin{z}$ is bounded by -1 and 1, and no specific interval is given in the problem motivation.

Therefore, values of $x$ for which the 4th-order Maclaurin polynomial approximate $\sin{x}$ with an error no greater than 0.001 should be $$(-0.001\times5!)^{1/5}\le x \le (0.001\times5!)^{1/5}$$ or $$-0.65438939 \le x \le 0.65438939$$ This is not the answer given in the solution manual. I'm clearly making a mistake, and I don't believe it's in my arithmetic. What bad assumption (or incorrect application of the error formula) am I using here?

  • $\begingroup$ Maybe they took into account that $|\sin x|\le |x|$, so it would be the $6$-th root? $\endgroup$
    – Bernard
    Apr 21 '19 at 23:46
  • $\begingroup$ Interesting. Could you elaborate? $\endgroup$ Apr 21 '19 at 23:47
  • 1
    $\begingroup$ We'd have $|R_n(x)| =\Bigl|\frac{x^5}{120}\Bigr||\sin c|$ for some $c$ between $0$ and $x$, so $|R_n(x)| \le\Bigl|\frac{x^5}{120}\Bigr||x|=\Bigl|\frac{x^6}{120}\Bigr|$, and ultimately $|x|^6\le 0.6$, whence $|x| \le\dots$ $\endgroup$
    – Bernard
    Apr 21 '19 at 23:55

As it turns out, when $x$ is sufficiently small, $\sin(x)$ is close enough to $x$ such that $$\frac{f^{(5)}(z)x^{5}}{5!} \approx \frac{x^6}{5!}$$

which gives an interval of $$-0.936410984 \le x \le0.936410984$$

This is pretty close to the answer, though for some reason the book takes the error up to the sixth order.

  • $\begingroup$ Oh! yes. The Maclaurin expansion for the cosine has no term of degree $5$, so the expansion at order $4$ is ipso facto the expansion at order $5$, and the error term can be calculated accordingly. $\endgroup$
    – Bernard
    Apr 22 '19 at 0:13
  • $\begingroup$ Okay, so then here's a gut-check: The Maclaurin expansion for sine has no even-ordered terms. So if I had to calculate the error of the $k$th order approximation of sine, then I would use $k$+2 in my error formula? $\endgroup$ Apr 22 '19 at 0:16
  • $\begingroup$ If $k$ is odd, of course. $\endgroup$
    – Bernard
    Apr 22 '19 at 9:13

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.