For a function $f: \mathbb R \to \mathbb R$ of locally bounded variation, when is $$\liminf_{e \to 0} V(f)[x, x+e]/e $$finite everywhere? Here $V(f)[a, b]$ denotes the total variation of the function on the interval $[a, b]$.

  • 1
    $\begingroup$ An obvious sufficient condition for this property is that $f$ be absolutely continuous with a locally bounded density. As for a necessary and sufficient condition, it is unclear in what terms you want such a characterization to be. I think it's possible that there are no (more or less) simple characterizations of this property, except for the trivial tautology: the property holds iff it holds. This property reminds me a bit the strong law of large numbers for not necessarily independent identically distributed random variables, for which I think no general nontrivial characterization exists. $\endgroup$ – Iosif Pinelis Jan 28 at 14:32
  • $\begingroup$ Sorry, what did you mean by density? $\endgroup$ – James Baxter Jan 28 at 14:34
  • 1
    $\begingroup$ By one of equivalent definitions of absolute continuity (en.wikipedia.org/wiki/Absolute_continuity), if $f$ is absolutely continuous, then the corresponding (signed) Lebesgue--Stieltjes measure $\mu_f$ is absolutely continuous with respect to the Lebesgue measure $\lambda$. What I mean then by the density is the Radon--Nikodym density $d\mu_f/d\lambda$. Can you answer the question about desired/expected terms of the characterization of the property? $\endgroup$ – Iosif Pinelis Jan 28 at 14:45
  • $\begingroup$ Hmm it’s hard to explain the full motivation, but I want something sort of like a weaker derivative for locally bounded variation functions. $\endgroup$ – James Baxter Jan 28 at 14:48
  • $\begingroup$ Because requiring a derivstive to exist at every point would be a little too extreme for the class of BV functions, I want some kind of “rough” measure of the pointwise change in the function, if that makes sense? $\endgroup$ – James Baxter Jan 28 at 14:49

By Locally BV, $Vf(a,b)$ is finite. Moreover $Vf(a,b) \ge Vf(c,d)$,whenever $(c,d) \subset (a,b)$. So you are right, its finite always.

Edit : After I read the question correctly.

consider $f(x) = 0$ everywhere except at $x = x_0$, where $f(x_0) = a \in \mathbb{R}$. $f$ is locally BV. So you are wrong. The limit is not always finite.

  • $\begingroup$ Right about what? $\endgroup$ – Wojowu Jan 28 at 12:25
  • $\begingroup$ always finite. I mean about being finite. $\endgroup$ – user102868 Jan 28 at 12:56
  • $\begingroup$ $V(f)[a,b]$ is finite, sure. But this is not what the question is about. $\endgroup$ – Wojowu Jan 28 at 13:10
  • $\begingroup$ @Wojowu : thanks for pointing. I have edited the answer, after reading the question correctly. $\endgroup$ – user102868 Jan 28 at 13:15
  • $\begingroup$ You are again answering the wrong question. I presume OP knows the limit is not always finite, the question is under what condition it is. $\endgroup$ – Wojowu Jan 28 at 13:48

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.