You probably should generate the series, first. My guess is that it is easier than you might think.

Results 1 to 6 of 6

- Aug 22nd 2008, 09:00 AM #1

- Joined
- Jun 2008
- Posts
- 53

## How big should n be?

Let p(x) be the taylor polynomial of degree n of the function

f(x) = log(1-x), about a = 0. How large should n be chosen to have

|f(x) - p(x)| <= 10^-4 ?

So I know how to go about this, but for Rn, what is f^(n+1) ?

Which derivative do I use? How do I know which to do?

Can someone give me the approximate n I'm looking for and the formula you used to get there, with details?

Thanks

- Aug 22nd 2008, 11:18 AM #2

- Joined
- Aug 2007
- From
- USA
- Posts
- 3,111
- Thanks
- 2

- Aug 22nd 2008, 02:03 PM #3

- Joined
- Jun 2008
- Posts
- 53

I'm sure I'm making it much harder.

f(x) = log(1-x), f(a)= 0

f'(x) = 1/(x-1), f'(a) = -1

f''(x) = 1/(x-1)^2, f''(a) = -1

So I'm thinking for f^(n+1) = -n! / (1-x)^(n+1)

is this right?

If so, |R(x)| <= |1/2|^(n+1)/ (n+1)! * (n!)/(1/2)^n+1

= 1 / n+1

right?

So n would need to be what, like 5000?

- Aug 22nd 2008, 02:45 PM #4

- Joined
- Aug 2007
- From
- USA
- Posts
- 3,111
- Thanks
- 2

- Aug 22nd 2008, 03:07 PM #5

- Joined
- Jun 2008
- Posts
- 53

- Aug 22nd 2008, 06:16 PM #6

- Joined
- Aug 2007
- From
- USA
- Posts
- 3,111
- Thanks
- 2

The anti-derivative thing and uniform convergence probably would not have been covered in an introductory course. Your series of derivative was fine. Consider that a hint that once you ahve paid your dues there will be cooler ways to go about it.

As far as bounds on the error, there must have been some discussion on the difference between alternating and non-alternating series. They ARE different animals. The result may be very similar, but getting there might take a different path.