Distribution of a random variable that is produced by another uniformly distributed random variable

260 Views Asked by At

Assume a simple two-step lottery: the reward (denoted by $Y$) is a random variable and realized as follows:

First: random variable $X$ is drawn from uniform distribution by interval $(a,c)$.

Second: After the realization of $X$, a player determines reward ($Y$) stochastically:

  • with probability $\alpha$: reward is $bX$ ($b$ is constant)

  • with probability $1-\alpha$: reward is $X$

Now, what is the distribution of the random reward $Y$?

If the solution is weighted average on both states: i.e. $Y=\alpha bX+(1-\alpha)X=(\alpha b+1-\alpha )X$

then we can say $Y$ is a linear rescaling transformation of $X$ and so $Y$ has the same distribution as $X$ has. But I am suspicious that $Y$ really is a weighted average of $X$ and $bX$

2

There are 2 best solutions below

5
On BEST ANSWER

Let $E$ denote the event that the reward is taken to be $bX$.

Then:$$P(Y\leq y)=P(E)P(Y\leq y\mid E)+P(E^c)P(Y\leq y\mid E^c)=$$$$\alpha P(bX\leq y\mid E)+(1-\alpha)P(X\leq y\mid E)$$ If moreover $1_E$ and $X$ are independent then this reduces to:$$P(Y\leq y)=\alpha P(bX\leq y)+(1-\alpha)P(X\leq y)\tag1$$

Can you take it from here?


Looking at it as weighted average as in your question the stochastic character of the choice is disregarded.

Actually we are dealing with:$$Y=1_EbX+(1-1_E)X\tag2$$where $1_E$ is a random variable (Bernoulli distribution with parameter $\alpha$).

This instead of $\alpha bX+(1-\alpha)X$.

3
On

The statement "$Y$ is a weighted average of $X$ and $bX$" is false.

What is true is: "the density (or CDF) of $Y$ is a weighted average of the densities (or CDF) of $X$ and $bX$".

This is known as a mixture.