Upperbound of the expected value of bounded variables

1k Views Asked by At

I have a general question about expected value.

If $C>0$ is a constant, and $(X_n)$ is a sequence of variables with $X_n<C$ for every $n\ge 1$, is it true that $E[X_n]\le C$ for every $n\ge 1$?

And similarly, if Y is a random variable, and $(X_n)$ is a sequence of variables with $X_n<Y$ for every $n\ge 1$, is it true that $E[X_n]\le E[Y]$ for every $n\ge 1$?

1

There are 1 best solutions below

0
On BEST ANSWER

By the definition of the expectation, if the density exists, then

$$E[X_n]=\int_{-\infty}^Cxf_X(x)\ dx\leq C\int_{-\infty}^{\infty}f(x)\ dx=C$$

because the support if the density $\subset (-\infty,C).$

If the density does not exist then the argumentation is the same but it starts with

$$E[X]=\int_{\infty}^C x\ dF_X.$$

For the second question, if $Y<X$ then $$E[X]<E[Y]$$ because every realizations of $X$ and $Y$ are in the same relationship.


The measure theoretical definition of the expectation, when the common probability space $[\Omega,\mathscr A,P]$ is given is

$$\int_{\Omega}X(\omega)\ dP.$$

The result is obvious if you take into account that the measurable function $X$ is changed to $C$ or a pointwise greater $Y$.