I think I need some help with this problem.

According to p.544 of Feller's An Introduction To Probability Theory. Vol II if $|a_k|\leq c_k$, $|b_k|\leq c_k$ then we have $|\prod_{i=1}^na_i -\prod_{i=1}^n b_i|\leq\sum_{k=1}^n (a_k\prod_{i=1}^{k-1}c_i-b_k\prod_{i=k+1}^n c_i)$

I could only prove $|a^n-b^n|=\bigg|\sum_{i=0}^{n-1}(a^{i+1}b^{n-(i+1)}-a^ib^{n-i})\bigg|=\bigg|\sum_{i=0}^{n-1}a^{i+1}b^{n-1-i}-\sum_{i=0}^{n-1}a^{i}b^{n-i}\bigg|\\=\bigg|(a-b)\sum_{i=0}^{n-1}a^ib^{n-1-i}\bigg|\leq n|a-b|c^{n-1}$

but I don't know how I can do the more general case.

  • $\begingroup$ What does $\sum_{k=1}^n c_1\cdot ... \cdot c_{k-1}a_k-b_kc_{k+1}\cdot ...\cdot c_n$ mean? Do you mean $\prod_{k=1}^n c_ka_k-\prod_{k=1}^nc_kb_k$? $\endgroup$ – R. Burton May 22 at 19:10
  • $\begingroup$ @R.Burton I changed it, hope its more clear now $\endgroup$ – Testname420 May 22 at 19:23
  • $\begingroup$ You may want to double check your formulas. $\endgroup$ – R. Burton May 22 at 19:39
  • $\begingroup$ The inequality is used on p. 544 in An Introduction to Probability Theory and Its Applications Vol2 from Feller $\endgroup$ – Testname420 May 22 at 19:41
  • $\begingroup$ Aah, that helps a lot. $\endgroup$ – R. Burton May 22 at 19:47

Assuming that the line "$|\alpha_1\cdots\alpha_n-\beta_1\cdots\beta_n|\leq\sum_{k=1}^n\gamma_1\cdots\gamma_{k-1}\alpha_k-\beta_k\gamma_{k+1}\cdots\gamma_n$" is indeed a mistake, and that it is meant to read $|\alpha_1\cdots\alpha_n-\beta_1\cdots\beta_n|\leq\sum_{k=1}^n|\gamma_1\cdots\gamma_{k-1}\alpha_k-\beta_k\gamma_{k+1}\cdots\gamma_n|$, we have...

$$\forall k,n\in\Bbb{N}.|a_k|\leq c_k\land |b_k|\leq c_k\implies\left\vert\prod_{i=1}^na_i -\prod_{i=1}^n b_i\right\vert\leq\sum_{k=1}^n \left\vert a_k\prod_{i=1}^{k-1}c_i-b_k\prod_{i=k+1}^n c_i\right\vert$$

Now, from this post, we know that...

$$\forall\mathbf{a},\mathbf{b}\in\Bbb{C}^n:\left(\forall k\leq n\in\Bbb{N}.|a_k|\leq 1\land |b_k|\leq 1\right).\left\vert\prod_{i=1}^na_i -\prod_{i=1}^n b_i\right\vert\leq\sum_{k=1}^n \left\vert a_k-b_k\right\vert$$

That $\forall k\leq n\in\Bbb{N}.|a_k|\leq 1\land |b_k|\leq 1$ is implied by context (if not, then disregard this answer), so the question becomes "how does Feller's sum differ?"

From $|\cdot|:\Bbb{C}\to[0,\infty)$, we know that $\forall k\in\Bbb{N}.|a_k|,|b_k|\geq 0$, whence...

$$\forall k\in\Bbb{N}.0\leq|a_k|,|b_k|\land|a_k|,|b_k|\leq c_k\implies 0\leq c_k$$

Thus, $\prod_{k\in K}c_k\geq0$ for all valid indexing sets $K$. Saying anything more than this will require more information on what $\mathbf{c}$ is. Suffice to say that if...

$$\exists x\in[1,\infty).\prod_{i=1}^{k-1}c_i=\prod_{i=k+1}^n c_i=x$$


$$\sum_{k=1}^n \left\vert a_k-b_k\right\vert\leq\sum_{k=1}^n \left\vert a_k\prod_{i=1}^{k-1}c_i-b_k\prod_{i=k+1}^n c_i\right\vert=x\sum_{k=1}^n \left\vert a_k-b_k\right\vert$$

  • $\begingroup$ Thanks you for your hard work $\endgroup$ – Testname420 May 24 at 15:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.