Expected value of $F(x)$.

369 Views Asked by At

$$F(X)=\frac{99}{1-X}\quad\text{and}\quad X\equiv Uniform[0,1]$$ What is the expected value of above distribution. I tried calculating but I find the expected value integral doesn't converge.

Is my calculation correct? What is the implication of expected value of integral not converging? Is the variance also not convergent?

1

There are 1 best solutions below

6
On BEST ANSWER

Indeed, it does not converge: $$ \mathbb{E}[F(X)] = \int_0^1 F(x)dx = \int_0^1 \frac{99}{1-x}dx = +\infty $$

This is alright, and can happen: not every random variable has a finite expectation! (For instance, look at the famous Cauchy distribution.)

What this means is exactly that: the expectation of the random variable $F(X)$ is infinite. (It is, at least, defined, since $F(X)\geq 0$: it just is infinite. The expectation of a Cauchy r.v. is not even defined, since a Cauchy r.v. takes both negative and positive values.)

A consequence? Indeed, then it does not have a variance either. This is because "having a variance implies having an expectation" (or, in fancier terms, $L_1$-integrable implies $L_2$-integrable for random variables). To show why, you can e.g. use Cauchy--Schwarz: $$ 0 \leq \mathbb{E}[\lvert Y\rvert ] = \mathbb{E}[\lvert Y\rvert\cdot 1] \leq \sqrt{\mathbb{E}[Y^2]\mathbb{E}[1]} = \sqrt{\mathbb{E}[Y^2]} $$ Thus, if $\mathbb{E}[Y^2] <\infty$, then $\mathbb{E}[\lvert Y\rvert ]<\infty$ as well, meaning that $Y$ is integrable (has a finite expectation).