Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Let $X_i \sim F_i$, where $F_i = G(F_{i-1})$ where there sequence begins with a known $F_0$ (and all $F_i$ are valid cdfs (monotonic increasing) defined over (0,1)).

I believe from other posts, that it is not enough to show that $\mathbb{E}X_0> \mathbb{E}X_{\infty}$ (where $X_{\infty}$ is the the random variable distributed by $F_i$ as $i \to \infty$, which we can assume is a Dirac mass function) to prove that $\mathbb{E}X_i$ decreases with $i$.

I believe the reason for this is that there could be a converging zig zag behavior, so that the iterations of $F_i$ could make the corresponding expectations go up a little, then down a little, while still converging to a fixed point.

My question is -- are there are properties of the $G$ operator that would allow you to say that there are no zigzags, and thus the above conditions are sufficient for proving the mean decreases with $i$?

Perhaps this is naive, but it would seem to me that most functions that someone like myself would encounter (read: a beginner) would not create zigzag behavior. Since G does not depend on n, it can't be raising a negative number to alternating powers or anything. And we can assume no trig operators.

Is it possible to say something like "$EX_0 > EX_{\infty}$, and the $G$ function only contains such-and-such type operations, therefore $EX_i$ decreases with $i$" ? And if so, what is "such-and-such" ?

share|improve this question
1  
Since $\mathbb{E}(X_i) = \int_0^1 (1-F_i(x)) \mathrm{d} x = \int_0^1 (1-G(F_{i-1}(x))) \mathrm{d} x$, if $G$ is such that $G(0)=0$, $G(1) = 1$ and $0 \le G(x) \le x$, then $\mathbb{E}(X_i) \ge \mathbb{E}(X_{i-1})$. –  Sasha Oct 14 '11 at 17:57
    
@Sasha: Sufficient, but not necessary, correct? –  Jand Oct 14 '11 at 18:08

1 Answer 1

up vote 1 down vote accepted

It is actually not that hard to find an example of $G(x)$ where the sequence $\mu_k = \mathbb{E}(X_k)$ is not monotonic. Here is one found with little help from Mathematica:

$$ G(x) = x \left( 1 +\frac{2429}{5002} (1-x) -\frac{83}{51} (1-x)^2 + \frac{113}{102} (1-x)^3\right) $$

enter image description here

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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