1. Limited time only! Sign up for a free 30min personal tutor trial with Chegg Tutors
    Dismiss Notice
Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Error Bounds for derivative estimation

  1. Sep 22, 2012 #1
    So we are estimated derivatives using a three point formula in class and the giving bounds for the error. I was given a table of x1=1.1, x2=1.2, x3=1.3, x4=1.4. I have to find an error bound to estimate the error for x=1.1 and 1.3 for f(x)=ex

    In class we did x=1.1 as follows

    error<=(h2/3)eg(x) for some g between (1.1, 1.3)
    error<=(h2/3)e1.3
    I understand all of this, except why it is between 1.1 and 1.3. I went to do the x3 error bound, and I wasn't sure what to choose for my upper bound. She added 2h to it, so would I use 1.5, even though it's not in my table?
     
  2. jcsd
  3. Sep 22, 2012 #2

    Ray Vickson

    User Avatar
    Science Advisor
    Homework Helper

    Do you mean ex? You can type that either as e^x or use the "X2" button in the menu at the top of the input panel. Also: you write "error<=(h2/3)eg(x)". That could mean several things (and I truly do not know which you intend). It could mean:
    [tex] (1) \frac{2}{3}h e g(x),\\
    (2) \frac{h^2}{3} e g(x),\\
    (3) \frac{2}{3} h e^{g x}.\\
    (4) \frac{h^2}{3} e^{g x}.
    [/tex]
    In plain text you could write the first as (2/3)*h* e*g(x), the second as (h^2 /3)*e*g(x), the third as (2/3)*h*e^(g x) and the fourth as (h^2 /3) * e^(g x); or, you could use the "X2" button to get (2/3)*h*e*g(x), (h2/3)*e*g(x), (2/3)*h*egx and (h2/3)*egx.

    RGV
     
  4. Sep 24, 2012 #3
    I'm sorry, I meant e^x, I copied it from somewhere else and didn't catch the format change. I'm sorry about that!
    I want to know (h^2)/3 * e^x, so the error was defined by h^2/3 * e^1.3
     
  5. Sep 24, 2012 #4

    Ray Vickson

    User Avatar
    Science Advisor
    Homework Helper

    The actual error will have the form (h^2)/3 * exp(y), for some y between 1.1 and 1.3, so--of course--the maximum possible error is obtained if we take the largest possible value of exp(y), which means taking the largest possible y; that is, y = 1.3. Is that what is bothering you?

    RGV
     
  6. Sep 25, 2012 #5
    No, It is when I move to evaluate it for x=1.3. I was given x values for 1.1, 1.2, 1.3, and 1.4. When I evaluate 1.3, what is the range of possible y values that I can consider? Does it stay between (1.1, 1.3) or do I move to include the points I used in my derivative estimation?
     
  7. Sep 25, 2012 #6

    Ray Vickson

    User Avatar
    Science Advisor
    Homework Helper

    It stays in the interval used in the derivative estimation. So, if you use a forward difference
    [tex] f'(x) \approx \frac{f(x+h) - f(x)}{h}[/tex] you would have y in [x,x+h]. If you use a central difference
    [tex] f'(x) \approx \frac{f(x+h) - f(x-h)}{2h}[/tex] you would have y in [x-h,x+h], etc. Didn't your textbook or course notes discuss this?

    RGV
     
Know someone interested in this topic? Share this thread via Reddit, Google+, Twitter, or Facebook