You are not logged in.
Pages: 1
I can't grasp the concept of Dirichlet's Test,
I've looked online, but I don't understand it. Can someone give me a simple example to explain it.
Thanks
edit: I think what is confusing me is this:
Am I supposed to prove that is true or do I just assume it is?
Last edited by woodoo (2007-02-11 19:56:17)
Offline
We have two sequences of real numbers {a[sub]n[/sub]} and {b[sub]n[/sub]} such that {a[sub]n[/sub]} is a decreasing positive sequence (a[sub]n[/sub] ≥ a[sub]n+1[/sub] > 0) which converges to 0 and the sum of countably many terms of {b[sub]n[/sub]} is bounded (|∑b[sub]n[/sub]| ≤ M for some constant M). Then
converges.
I will use the problem you posted in your other thread to give a simple example:
Let {a[sub]n[/sub]} = 1/n and {b[sub]n[/sub]} = sin n. Clearly, 1/n ≥ 1/(n + 1) > 0 and lim(1/n) = 0, so that {a[sub]n[/sub]} satisfies the conditions of Dirichlet's test. Since |sin x| ≤ 1 for any real x, we have that
for every positive integer N. Then {b[sub]n[/sub]} satisfies the conditions for Dirichlet's test. Thus
converges.
Last edited by Zhylliolom (2007-02-11 20:12:38)
Offline
Oh I understand now, I thought that {bn} was going to have to equal (sin n)/n, I didn't realize you were supposed to break down (sin n)/n into {an} = 1/n and {bn} = sin n to represent
Thank you Zhylliolom
Last edited by woodoo (2007-02-11 20:16:46)
Offline
Cool. I learned something, too!
"The physicists defer only to mathematicians, and the mathematicians defer only to God ..." - Leon M. Lederman
Offline
I'm not sure I get this either.
If we used the same reasoning as we did for sin n/n, but replaced sin n with 1, then according to Dirichlet's test it would converge, because all of the same reasoning still holds.
1≤1 for all x, so
And yet, it's been proven that 1/n diverges. What am I doing wrong?
Why did the vector cross the road?
It wanted to be normal.
Offline
I'm not sure I get this either.
If we used the same reasoning as we did for sin n/n, but replaced sin n with 1, then according to Dirichlet's test it would converge, because all of the same reasoning still holds.
1≤1 for all x, so
And yet, it's been proven that 1/n diverges. What am I doing wrong?
Great point!
And I don't think f(N) is bounded because |f(N)|≤N.
X'(y-Xβ)=0
Offline
I'm not sure I get this either.
If we used the same reasoning as we did for sin n/n, but replaced sin n with 1, then according to Dirichlet's test it would converge, because all of the same reasoning still holds.
1≤1 for all x, so
And yet, it's been proven that 1/n diverges. What am I doing wrong?
I think a minor mistake/typo has been made: for Dirichlet's test, you need
for every M, and for N a constant. (note the change in limit of the summation from N to M)This is the case for b[sub]n[/sub] = sin(n), but not for b[sub]n[/sub]n = 1.
Last edited by Dross (2007-02-12 02:18:17)
Bad speling makes me [sic]
Offline
Ah, that makes a bit more sense. But then again, surely adding up lots of sin n's would eventually get you to a number greater than any M you choose? How does that bit work?
Why did the vector cross the road?
It wanted to be normal.
Offline
That is hard
X'(y-Xβ)=0
Offline
Ah, that makes a bit more sense. But then again, surely adding up lots of sin n's would eventually get you to a number greater than any M you choose? How does that bit work?
Some of the sin(n) terms will be positive, and some will be negative (remember you only take the modulo after adding all the terms), so they will cancel each other out.
Bad speling makes me [sic]
Offline
Gah! Of course, I completely forgot about sin x being sometimes negative. Now it makes perfect sense.
Why did the vector cross the road?
It wanted to be normal.
Offline
So if Zhylliolom right? Because you guys are confusing me
Offline
Zhylliolom is nearly right. The only change you need to his explanation is to use Dross's formula instead of his, when talking about a[sub]n[/sub].
Sorry, I kind of hijacked your thread a bit there.
Why did the vector cross the road?
It wanted to be normal.
Offline
mathsyperson wrote:Ah, that makes a bit more sense. But then again, surely adding up lots of sin n's would eventually get you to a number greater than any M you choose? How does that bit work?
Some of the sin(n) terms will be positive, and some will be negative (remember you only take the modulo after adding all the terms), so they will cancel each other out.
That is not rigorous. Look at this sequence:
2, -1, 2, -2, 2, -0.5,...
Are you sure of convergence?
X'(y-Xβ)=0
Offline
That is not rigorous. Look at this sequence:
2, -1, 2, -2, 2, -0.5,...
Are you sure of convergence?
Oh, it's not rigorous at all - more a view to see that it's possible. I just can't remember how to prove it at the moment, is all.
Bad speling makes me [sic]
Offline
Pages: 1