Interchange Order of Integrals

827 Views Asked by At

Can someone explain the last step in this process. Specifically, how do you get the new limits of integration?

Expected Value Definition: $E[Y] = \int_0^\infty{P\{Y \ge y\} \, dy}$

Expand: $E[Y] = \int_0^\infty{\int_y^\infty{f_Y(x) \, dx} \, dy}$

Clarify: $E[Y] = \int_{y=0}^{y=\infty}{\left(\int_{x=y}^{x=\infty}{f_Y(x) \, dx}\right) \, dy}$

Interchange order of integration: $E[Y] = \int_{x=0}^{x=\infty}{\left(\int_{y=0}^{y=x}{dy} \right) f_Y(x) \, dx}$

1

There are 1 best solutions below

0
On BEST ANSWER

Assume that $f_Y$, the pdf of $Y\ge 0$, exists. In this case the expected value is

$$E[Y]=\int_{0}^{+\infty}x\ f_Y(x)\ dx$$

Also, assume that the integral above is finite.

It is obvious that $x$ can be written as

$$x=\int_0^x\ 1\ dy, \ x\ge 0.$$

With this, the expected value is

$$E[Y]=\int_{0}^{+\infty}\int_0^x\ 1\ dy\ f_Y(x)\ dx.$$

Since the expectation exists we can use Fubini' theorem so we can compute this integral as a double integral.

$$E[Y]=\iint_A1\cdot d\mu(x,y)$$ where $$A=\{(x,y): 0\le y\le x\ \text{ and } 0 \le x < \infty \}.$$ and $\mu$ is the product measure of the probability measure generated by $f_Y$ and the Lebesgue measure. (For a rectangle $r=[a,b]\times [c,d]$, the measure $\mu(r)=(b-a)\int_c^df_Y(x)\ dx$.)

The domain of the integration can be visualized:

enter image description here

Considering the definition of $\mu$ and the shape of the domain, and referring to Fubini's theorem, the double integral above can be written as

$$\int_{0}^{+\infty}\int_{y}^{+\infty}\ 1\ f_Y(x)\ dx\ dy=$$ $$=\int_{0}^{+\infty}P(Y\ge y)\ dy.$$