Finite expectation implies absolutely convergent?

83 Views Asked by At

This is a statement from Shiryaev's Probability that I wonder if somebody can explain to me. He starts with a definition:

Definition: We say that the expectation of $\xi$ is finite if $E \ {\xi^+} < \infty$ and $E \ {\xi^-} < \infty$.

($\xi^+ =max(\xi,0), \xi^-=-min(\xi,0)$).

Then comes the statement that I'm not sure about:

Since $|\xi|={\xi^+}+{\xi^-}$ the finiteness of $E\ \xi$, or $|E \ \xi |<\infty$ is equivalent to $E\ |\xi| <\infty$

So if I know that $E\ \xi < \infty $ then $E \ |\xi| < \infty$?

I find this a bit counter intuitive; I mean in general $\int f dx < \infty $ doesn't imply $\int |f|dx < \infty$. Why is it different in this case?