I'm using Resnick's "A probability path" and I'm bit confused with his notation (particularly regarding $\uparrow$ and $\downarrow$ )when proving the continuity of the measure P for monotone sequences (page 31) and when proving Fatou's lemma (page 32).

1. Continuity of P for monotone sequences

If $A_n \uparrow A$, where $A_n \in \mathcal{B}$, then $P(A_n) \uparrow P(A)$

In proposition 1.4.1. (page 8), the author talks about monotone sequences and defines $A_n \uparrow $ as indicating that $A_n$ is a monotone non-decreasing sequence. In that case, then we can define the limit of $A_n \uparrow $ as:

$$ \lim_{n \rightarrow \infty} A_n = \cup_{n=1}^\infty A_n $$

So I can interpret $A_n \uparrow A$ as saying that the sequence $A_n$ is non-decreasing and $A = \cup_{n=1}^\infty A_n$.

My problem is I'm not sure what $P(A_n) \uparrow P(A)$ means. Is it a regular limit, but the $\uparrow$ indicates that $P(A_n)$ is non-decreasing? Or does it indicate a $lim inf$ or some other convergence concept?

Concretely, I'm confused by the following step:

First he constructs a disjoint sequence of events

$$B_1 = A_1, B_2 = A_2 \setminus A_1, ..., B_n = A_n \setminus A_{n-1}, ... $$

and so $ \cup_{i=1}^n B_i = A_n, \cup_{i=1}^n B_i = \cup_{i} A_i ) = A$

And then:

$$ P(A) = P(\cup_{i=1}^\infty P(B_i) B_i) = \sum_{i=1}^\infty P(B_i) = \lim_{n \rightarrow \infty} \uparrow \sum_{i=1}^n P(B_i) $$

I'm confused by that last equality and what $\lim_{n \rightarrow \infty} \uparrow \sum_{i=1}^n P(B_i) $ means. I'm guessing I can't say that

$$ \sum_{i=1}^\infty P(B_i) = \lim_{n \rightarrow \infty} \sum_{i=1}^n P(B_i) $$

because then I would be able to conclude that $P(A_n) \rightarrow P(A)$, which seems stronger that what the property says, but I don't get why it wouldn't be true.

This answer might be relevant, though it doesn't use the $\uparrow $ notation


2 Answers 2


You can view $P(A_n) \uparrow P(A)$ as saying that $P(A_n)$ converges to $P(A)$ from below. Resnick writes $P(A_n) \uparrow P(A)$ as shorthand for saying both that

  1. $\lim_{n \to \infty} P(A_n) = P(A)$, and
  2. $P(A_n) \leq P(A_{n+1})$ (the sequence is non-decreasing).

In light of item 2 above, you can replace 1 by something like, $P(A) = \sup_n P(A_n)$ or $P(A) = \liminf_{n \to \infty} P(A_n)$, et cetera.

Similarly, the "$\lim_{n\to\infty} \uparrow$" notation signifies a limit from below. In the case of $\sum_{i=1}^\infty P(B_i)$, it's just saying that a series of non-negative terms is the limit of the non-decreasing sequence of its partial sums.

  • $\begingroup$ Thanks! That cleared up stuff $\endgroup$
    – cd98
    Commented Aug 26, 2020 at 16:44

I mostly agree with Artem's answer except for the comment that the $\lim_{n\to\infty} \uparrow$ notation signifies a limit from below. If this were true then $\lim_{n\to\infty} \downarrow (\bigcup_{k \ge n} A_k)$ (on page 33, in the proof of Fatou's lemma) would signify a limit from above, which does not make sense. It seems that the down arrow in my example is merely signifying that $\{\bigcup_{k \ge n} A_k\}$ is a monotone nonincreasing sequence and therefore the limit is well-defined using Proposition 1.4.1. It is a little superfluous and can be ignored. In the original question, $\lim_{n \rightarrow \infty} \uparrow \sum_{i=1}^n P(B_i)$ is the same as $\lim_{n \rightarrow \infty} \sum_{i=1}^n P(B_i)$. After all, $\{\sum_{i=1}^n P(B_i)\}$ is a monotone nondecreasing sequence that is also bounded (since $B_1, B_2, \dotsc$ are disjoint and the sum of the probabilities assigned to them is at most $1$) and the limit of such a sequence always exists using the monotone convergence theorem.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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