MHB Comparing Difference Quotients for Approximating $f'''(x)$

  • Thread starter Thread starter evinda
  • Start date Start date
  • Tags Tags
    Difference
AI Thread Summary
The discussion focuses on the accuracy of two difference quotients for approximating the third derivative, $f'''(x)$. The first difference quotient has an error bound of $\frac{22}{4} h ||f^{(4)}||_{\infty}$, while the second has a smaller error bound of $\frac{3}{4} h ||f^{(4)}||_{\infty}$. This suggests that the second difference quotient is a more accurate approximation due to its smaller constant. Participants discuss the need for a formal justification of this conclusion, with suggestions to express the error terms in a sharper form. Overall, the conversation emphasizes the importance of error analysis in determining the accuracy of numerical approximations.
evinda
Gold Member
MHB
Messages
3,741
Reaction score
0
Hello! (Wave)

I have to show that the following difference quotients are approximations of $f'''(x)$.

$$\frac{f(x+3h)-3f(x+2h)+3f(x+h)-f(x)}{h^3} \\ \frac{f(x+2h)-2f(x+h)+2f(x-h)-f(x-2h)}{2h^3}$$

Which approximation is more accurate? Justify your answer.I found the Taylor expansion of $f(x+3h) , f(x+2h), f(x+h)$ and found that

$$\left| \frac{f(x+3h)-3f(x+2h)+3f(x+h)-f(x)}{h^3}-f'''(x) \right| \leq \frac{22}{4} h ||f^{(4)}||_{\infty}$$

Have we shown now that $\frac{f(x+3h)-3f(x+2h)+3f(x+h)-f(x)}{h^3}$ is an approximation of $f'''(x)$?
Or do we have to show that the above tends to $0$ ?

Similarly, I found that

$$\left| \frac{f(x+2h)-2f(x+h)+2f(x-h)-f(x-2h)}{2h^3}-f'''(x) \right| \leq \frac{3}{4} h ||f^{(4)}||_{\infty}$$

The second difference quotient is a better approximation because of the smaller constant, right ?

But how could we justify it formally? (Thinking)
 
Mathematics news on Phys.org
evinda said:
The second difference quotient is a better approximation because of the smaller constant, right ?

But how could we justify it formally? (Thinking)

Hey evinda! (Smile)

I think that what you have is already sufficiently formal.

The only thing I can think of to improve it, is to set the expressions equal to $\frac{22}{4} h f^{(4)}(x+\theta h)$ respectively $\frac{3}{4} h f^{(4)}(x+\xi h)$, where $0\le\theta\le 1$ and $0\le\xi\le 1$.
I'd consider that sharper than giving an upper bound. (Thinking)
 
I like Serena said:
Hey evinda! (Smile)

I think that what you have is already sufficiently formal.

The only thing I can think of to improve it, is to set the expressions equal to $\frac{22}{4} h f^{(4)}(x+\theta h)$ respectively $\frac{3}{4} h f^{(4)}(x+\xi h)$, where $0\le\theta\le 1$ and $0\le\xi\le 1$.
I'd consider that sharper than giving an upper bound. (Thinking)

Nice... Thanks a lot! (Smile)
 
Seemingly by some mathematical coincidence, a hexagon of sides 2,2,7,7, 11, and 11 can be inscribed in a circle of radius 7. The other day I saw a math problem on line, which they said came from a Polish Olympiad, where you compute the length x of the 3rd side which is the same as the radius, so that the sides of length 2,x, and 11 are inscribed on the arc of a semi-circle. The law of cosines applied twice gives the answer for x of exactly 7, but the arithmetic is so complex that the...
Just chatting with my son about Maths and he casually mentioned that 0 would be the midpoint of the number line from -inf to +inf. I wondered whether it wouldn’t be more accurate to say there is no single midpoint. Couldn’t you make an argument that any real number is exactly halfway between -inf and +inf?
Back
Top