Can I simply convert an integral into an expectation by "declaring" that the variable is a random variable?

252 Views Asked by At

I know that

$\int_{a}^{b}g(x)dx = (b-a)\int_{a}^{b}g(x)\frac{1}{b-a}dx = (b-a)E[g(X)]$

where $X$ has a $Uniform(a,b)$ distribution.

However, I don't understand how the second step turns the integral into an expected value. It's as if I can simply "declare" that $X$ is a random variable. Somehow, it seems like I'm hand-waving some magic into the equation, and I wonder if I am missing a step.

1

There are 1 best solutions below

0
On BEST ANSWER

As mentioned in the comments, this works because it follows from the definition of expectation of a continuous random variable.

Consider a continuous random variable $X$ defined on some interval $I$ with some distribution with pdf $f_X(x)$. Then,

$$\Bbb E[g(X)]:=\int_I g(x)f_X(x)~\mathrm dx$$

The question here corresponds to $X\sim\mathrm{Unif}(a,b)$ for which $f_X(x)=\frac 1{b-a}$ for $x\in [a,b]$ and $0$ otherwise, so we have,

$$\Bbb E[X]:=\int_a^b g(x)\frac 1{b-a}~\mathrm dx\implies (b-a)\Bbb E[X]=\int_a^b g(x)~\mathrm dx$$