Can anyone help me with my big problem about real analysis and measure theory. In Mathematics are a lot of theorem which are proved with the help of the notion $\mu$-almost everywhere . I know that $\mu$ is a measure, but I want to know when can I apply the theory of $\mu$-almost everywhere and when I cannot. Can you give me some example please.

Thanks :)

  • 1
    $\begingroup$ A property of elements of a measure space holds $\mu$-almost everywhere if there is a measurable set $N$ with measure zero such that the property holds for all elements of the measure space that are not in $N$. $\endgroup$ – Michael Greinecker Jan 28 '13 at 13:27

Well, to make things more messy I would say that $\mu$-a.e. results are $\mu$-relatively strong. Namely, such results do not hold on sets of very small important (of $\mu$ measure zero). It all depends on the measure $\mu$ in the end, and on how do you rely upon it. Namely, measures are weights that tell you what is more important and what is less.

If you know that

a bounded function $f:[0,1]\to\Bbb R$ is Riemann integrable iff the set of discontinuities $D(f)$ is of the Lebesgue measure zero

then it gives you a pretty nice characterization of those functions that are Riemann integrable. Indeed, you have both a cool intuition that such functions does not have to be very weird (their set of weirdness $D(f)$ is "small enough") and you have a working criteria to check integrability in all particular cases.

If you know that

the Poisson process has a finite number of jumps $\mathsf P$-a.e. (where $\mathsf P$ is a probability measure)

you will think, that simulating a trajectory of such process with exact random number generator will be successful in your whole entire life.

However, in case you are able to show that something holds $\delta_x$-a.e. where $\delta_x$ is a Dirac measure (a point mass) at a single point $x$, this is not a very useful result usually. Indeed, it tells you that some property holds at a point $x$ and does not give you any further information. Agree?

Well, as I mentioned above it's all relative and in fact some single point $x$ may be more important for you then the rest of the state space. For example,

the set $A$ is visited by a $\psi$-irreducible Markov Chain at least once regardless of the initial condition iff $\psi(A)>0$.

Here the measure $\psi$ tells us which sets are "big enough" w.r.t. the dynamics of a Markov Chain. Namely, which sets, no matter where do you start from, you'll visit with a positive probability. Such sets however do not have to be "big" in our usual understanding. Namely, it may happen naturally happen for some probabilistic models that $\psi(\{0\}) = 1$ and $\psi(\Bbb R\setminus \{0\}) = 0$. Does it mean that the singleton set $\{0\}$ is bigger than $\Bbb R\setminus \{0\}$? Relatively, yes - under these particular conditions the behavior of a process at $\{0\}$ determines the whole asymptotic behavior of it.

Hope, these examples help.

  • $\begingroup$ To be strictly correct, you'll want to change "the function $f:[0,1]\to\Bbb R$ is Riemann integrable iff" to "the bounded function $f:[0,1]\to\Bbb R$ is Riemann integrable iff". $\endgroup$ – Dave L. Renfro Jan 28 '13 at 17:07
  • $\begingroup$ @DaveL.Renfro: thanks for the typo $\endgroup$ – Ilya Jan 28 '13 at 18:37

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.