切比雪夫不等式概率证明步骤3疑问：为何E[I_{{X²>1}}] ≤ E[X²]

Great question! Let's break this down into straightforward, intuitive steps—this inequality hinges on two core concepts: how indicator functions work, and a basic property of expectation.

Step 1: What is the indicator function here? #

First, let's demystify $\mathbf{I}_{{ X^2 >1}}$. It’s a simple 0-1 random variable:

• For any outcome where $X^2 > 1$, the indicator equals 1.
• For any outcome where $X^2 \leq 1$, the indicator equals 0.

Step 2: Compare the indicator to $X^2$ for every possible outcome #

Now, let’s look at how these two values stack up for every scenario:

• When $X^2 > 1$: The indicator is 1, and since $X^2$ is greater than 1 here, $1 \leq X^2$. So $\mathbf{I}_{{ X^2 >1}} = 1 \leq X^2$.
• When $X^2 \leq 1$: The indicator is 0, and since squares are always non-negative ($X^2 \geq 0$), $0 \leq X^2$. So $\mathbf{I}_{{ X^2 >1}} = 0 \leq X^2$.

In plain terms: no matter what outcome we’re looking at, the indicator function is always less than or equal to $X^2$.

Step 3: Apply the monotonicity of expectation #

A key rule of expectation is monotonicity: if random variable $Y$ is less than or equal to random variable $Z$ for every possible outcome (or almost all, in formal measure theory terms), then $\mathbb{E}[Y] \leq \mathbb{E}[Z]$.

Since we just proved $\mathbf{I}{{ X^2 >1}} \leq X^2$ holds everywhere, applying this rule directly gives us the inequality you’re curious about:
$$\mathbb{E}\left[\mathbf{I}{{ X^2 >1}} \right] \leq \mathbb{E}\left[X^2\right]$$

Bonus: Link to Markov's Inequality #

This is actually a direct special case of Markov's Inequality, which says for any non-negative random variable $Y$ and positive constant $a$, $\mathbb{P}(Y > a) \leq \frac{\mathbb{E}[Y]}{a}$. Here, $Y = X^2$ and $a=1$, so $\mathbb{P}(X^2 >1) = \mathbb{E}[\mathbf{I}_{{X^2>1}}] \leq \frac{\mathbb{E}[X^2]}{1}$—exactly the inequality in question.

内容的提问来源于stack exchange，提问作者stollenm