Talk:Fatou's lemma

Where in the two given proofs is the measurability of f verified?Lost-n-translation (talk) 23:56, 22 March 2012 (UTC)

It follows immediatly because the limit inferior of measurable functions is measurable. 94.212.65.238 (talk) 12:48, 20 October 2013 (UTC)

Simplest version of lemma?

It seems like there is a much simpler version of this result that would be a more natural starting point -- the expectation of a minimum is less than the minimum of the expectations.

$<min_{i}x_{i}>\leq min_{i}<x_{i}>$

Proofs unreadable in the mobile version

The proofs in section Standard_statement_of_Fatou's_lemma cannot be read in the mobile version of this article because the div's will not open if clicked. --151.75.40.83 (talk) 12:01, 26 November 2015 (UTC)

Hmm... Everything looks fine for me. Are you using an up-to-date browser? StrokeOfMidnight (talk) 19:47, 16 November 2017 (UTC)

$g_{n}$ undefined

Starting in Remark 3, $g_{n}$ is referenced often but is never defined. I believe it should be $g_{n}=\inf\{f_{n},f_{n+1},...\}$, as in Zigmond's Measure and Integral.

Thank you so much for pointing this out. $g_{n}$ is defined exactly like you say, but I deleted that definition recently by accident. Everything else makes sense in this section, I hope? StrokeOfMidnight (talk) 19:39, 16 November 2017 (UTC)

Multiple issues with the proof

A few issues: First, the proof now defines $g_{n}$ but does not define $g$ before using it. Secondly, why are we requiring that the functions be Borel-measurable? Isn't Lebesgue-measurable sufficient? Finally, there are two sections in a row labeled "proof". What do you think of relabeling the second such section as "Alternative proof using the Monotone Convergence Theorem" or something like that?Ceresly (talk) 17:38, 29 December 2018 (UTC)

Ceresly: Excellent point: Lebesgue-measurability is enough. By the way, the article on Monotone Convergence Theorem has the exact same problem.

Which $g$ are you referring to?

It's been a while since I've noticed that multiple sections have been called "Proof", which is ugly misleading. One section, for example, should probably be renamed to "Proof of Lemma 1", and "Independent proof" should become something like "Proof independent of Monotone Convergence Theorem".

I've been trying to make this article as similar to the MCT one as possible, but I don't have time for this at the moment, and I don't want to rush. Therefore if someone else picks up the slack, that will be great. StrokeOfMidnight (talk)

StrokeOfMidnight: Thanks for the reply. Here is the "offending" section... it's in step one of the first proof:

"Since $f\leq g$, we have

$SF(f)\subseteq SF(g).$" Ceresly (talk) 13:46, 30 December 2018 (UTC)

Ceresly: Remark 5 says "let the functions $f,g:X\to [0,+\infty ]$ be ...", so both $f$ and $g$ are assumed to be arbitrary here, and no definition should be expected, although it might still make sense to name them, say, $g_{1}$ and $g_{2}$ to avoid confusion. StrokeOfMidnight (talk)

StrokeOfMidnight: Oh now I understand. I didn't realize that that was a proof of Remark 5, not a proof of Fatou's Lemma. Ceresly (talk) 22:07, 30 December 2018 (UTC)

Ceresly: I have to retract my previous answer: everything is correct on the measurability side. By definition, $f$ is Lebesgue-measurable iff the set $\{f\geq y\}$ is Lebesgue-measurable, for every $y\in [-\infty ,+\infty ].$ The intervals $[y,+\infty )$ generate the Borel $\sigma$-algebra. StrokeOfMidnight (talk)

StrokeOfMidnight: Not sure I understand. The hypothesis (the way it's written now) is that each function has to be Borel-measurable. I'm just saying that that it unnecessarily restrictive; Lebesgue-measurable would suffice. There are functions that are L-measurable but not B-measurable (e.g. the indicator function of an unmeasurable set, I think), and Fatou's Lemma would still apply to a sequence of functions like that. But I think maybe I've misunderstood you. Ceresly (talk) 16:35, 6 January 2019 (UTC)

Ceresly: Fatou's lemma (including its WP version) doesn't talk about Borel measurability at all. Instead, the lemma talks exclusively about measurability according to Lebesgue, and this is properly reflected in the article. Indeed, since closed intervals (or open ones, for that matter) generate the Borel $\sigma$-algebra, the above definition (see my previous answer) is equivalent to: $f$ is Lebesgue-measurable iff, for every Borel set $S\subseteq [-\infty ,+\infty ],$ the inverse image $f^{-1}(S)$ is Lebesgue-measurable. Here, $f:X\to [-\infty ,+\infty ],$ and there is a measurable space $(X,\Sigma ,\mu )$ on one side and the Borel $\sigma$-algebra ${\mathcal {B}}$ on the other. Hence, $f$ is $(\Sigma ,{\mathcal {B}})$-measurable. In retrospective, I think this notation might be a bit misleading. (By the way, Fatou's lemma only cares about the non-negative semiaxis $[0,+\infty ]$). StrokeOfMidnight (talk) 19:59, 6 January 2019 (UTC)

StrokeOfMidnight: Are you saying that "$f$ is $(\Sigma ,{\mathcal {B}})$-measurable" should be read as "$f$ is Lebesgue-measurable"? That seems to be what you're saying. That would be fine if that's what you're saying; I guess I've never seen that notation before. Ceresly (talk) 22:10, 6 January 2019 (UTC)

Ceresly: Yes, now we seem to be consistent. And, yes, I probably should explain this notation in the article. StrokeOfMidnight (talk)