How to Derive the Inequality on Page 36 in the Proof of Lemma 11.3?

In summary, the authors assume that $g$ is equal to $1$ on an inner interval and smoothly decreases to $0$ in the region between this interval and the target region $T$. They also assume that $f$ vanishes at some point in $T$. By using the mean value theorem, it can be shown that $|f'(x)g(x)+f(x)g'(x)|$ is bounded by the product of the supremum of $f'$ and $g'$ in $T$ and an additional term involving $\ell$.
  • #1
Alone
60
0
I am trying to see how to derive the following inequality on page 36 in the proof of Lemma 11.3: https://arxiv.org/pdf/math/0412040.pdf

I.e, of:
$$\| fg \|_{Lip} \le \bigg(1+\ell \sup_{t\in T} |g'(t)|\bigg)\sup_{t\in T}|f'(t)| , \ \ supp \ f(1-g)\subset S^c$$

My thoughts about how to show this, first let's write:
$$|f\cdot g (x)-f\cdot g(y)|/|x-y|$$
For $x\ne y$ and $x,y\in \mathbb{R}$; I think I should estimate the derivative of $f\cdot g$ which is $f'g+fg'$, but I don't see how exactly to use it here.

I hope this question is in the right level for this site.

Cheers!
 
Physics news on Phys.org
  • #2
Alan said:
I am trying to see how to derive the following inequality on page 36 in the proof of Lemma 11.3: https://arxiv.org/pdf/math/0412040.pdf

I.e, of:
$$\| fg \|_{Lip} \le \bigg(1+\ell \sup_{t\in T} |g'(t)|\bigg)\sup_{t\in T}|f'(t)| , \ \ supp \ f(1-g)\subset S^c$$

My thoughts about how to show this, first let's write:
$$|f\cdot g (x)-f\cdot g(y)|/|x-y|$$
For $x\ne y$ and $x,y\in \mathbb{R}$; I think I should estimate the derivative of $f\cdot g$ which is $f'g+fg'$, but I don't see how exactly to use it here.
You are right, you need to estimate $\sup |f'(x)g(x)+f(x)g'(x)|$. You only need to deal with the case $x\in T$, because $g$ (and therefore also $g'$) vanishes outside $T$. So you need to know how large $|f|$ and $|g|$ can be in the interval $T$.

The authors seem to assume that $g$ is equal to $1$ on the inner interval $I$, and then smoothly decreases to $0$ in the region between $I$ and $T$. They don't actually say so, but I think they are assuming that $g$ always lies between $0$ and $1$, so that $\sup_{t\in T}|g(t)| = 1$.

They also assume ("without loss of generality", though I don't see why) that $f$ vanishes at some point of $T$. Say $f(s) = 0$, where $s$ is some point of $T$. Every other point of $T$ is within distance $\ell$ of $s$, and it follows (by the mean value theorem) that $|f(x)| \leqslant \ell\sup_{t\in T} |f'(t)|$ for all $x$ in $T$.

Therefore, for all $x$ in $T$, $$|f'(x)g(x)+f(x)g'(x)| \leqslant \sup_{t\in T}|f'(t)| \sup_{t\in T}|g(t)|+ \sup_{t\in T}|f(t)|\sup_{t\in T}|g'(t)| \leqslant \sup_{t\in T}|f'(t)|\cdot 1 + \ell\sup_{t\in T}|f'(t)| \sup_{t\in T}|g'(t)|,$$ which is what you wanted to prove.
 

FAQ: How to Derive the Inequality on Page 36 in the Proof of Lemma 11.3?

What is the purpose of deriving an inequality from the paper "A CLT for a Band Matrix Model" by Anderson and Zeitoun?

The purpose of deriving an inequality from this paper is to establish a bound on the error term in the central limit theorem for a specific type of band matrix model. This can help in understanding the accuracy and reliability of the central limit theorem in this context.

How is the inequality derived in the paper?

The inequality is derived using a combination of analytical techniques and mathematical manipulations. The authors use various properties of band matrices, such as their eigenvalues and spectral norms, to derive the final inequality.

What is the significance of this inequality in relation to the central limit theorem?

This inequality provides a quantitative measure of the accuracy of the central limit theorem for the specific band matrix model studied in the paper. It shows that the error term in the central limit theorem is bounded by a certain value, which can help in understanding the behavior of the model and its deviations from the normal distribution.

Can this inequality be generalized to other band matrix models?

While the inequality is specific to the band matrix model studied in the paper, the techniques used for its derivation may be applicable to other similar models. However, the specific bounds and constants may differ depending on the properties of the particular model.

How does this inequality contribute to the field of statistics and mathematics?

This inequality adds to the existing body of knowledge on band matrix models and their central limit theorems. It provides a deeper understanding of the behavior and accuracy of these models, and may also have practical applications in various statistical and mathematical analyses that involve band matrices.

Similar threads

Replies
24
Views
3K
Replies
3
Views
2K
3
Replies
100
Views
9K
Replies
3
Views
3K
Replies
11
Views
2K
Back
Top