If $X$ is a discrete and $Y$ is a continuous random variable then what can we say about the distribution of $X+Y$? Is it continuous or is it mixed?

What about the product $XY$?


Suppose $X$ assumes values $k \in K$ with discrete distribution $(p_k)_{k \in K}$, where $K$ is a countable set, and $Y$ assumes values in $\mathbb R$ with density $f_Y$ and CDF $F_Y$.

Let $Z = X + Y$. We have $$ \mathbb P( Z \leq z) = \mathbb P(X + Y \leq z) = \sum_{k \in K} \mathbb P(Y \leq z - X \mid X = k) \mathbb P(X = k) = \sum_{k \in K} F_Y(z-k) p_k,$$ which can be differentiated to obtain a density function for $Z$ given by $$ f_Z(z) = \sum_{k \in K} f_Y(z-k) p_k.$$

Now let $R = X Y$ and assume $p_0 = 0$. Then $$ \mathbb P(R \leq r) = \mathbb P(X Y \leq r) = \sum_{k \in K} \mathbb P(Y \leq r/X) \mathbb P(X= k) = \sum_{k \in K} F_Y(r/k) p_k,$$ which again can be differentiated to obtain a density function.

However if $p_0 > 0$, then $\mathbb P(X Y = 0) \geq \mathbb P(X = 0) = p_0 > 0$, which shows that in this case $XY$ has an atom at 0.


Let $X$ be a discrete random variable with probability mass function $p_X : \mathcal{X} \to [0,1]$, where $\mathcal{X}$ is a discrete set (possibly countably infinite). Random variable $X$ can be thought of as a continuous random variable with the following probability density function

$$f_X (x) = \sum_{x_k \in \mathcal{X}} p_X (x_k) \, \delta (x - x_k)$$

where $\delta$ is the Dirac delta function.

If $Y$ is a continuous random variable, then $Z := X+Y$ is a hybrid random variable. As we know the probability density functions of $X$ and $Y$, we can compute the probability density function of $Z$. Assuming that $X$ and $Y$ are independent, the probability density function of $Z$ is given by the convolution of the probability density functions $f_X$ and $f_Y$

$$f_Z (z) = \sum_{x_k \in \mathcal{X}} p_X (x_k) \, f_Y (z - x_k)$$

  • $\begingroup$ Why the downvote? $\endgroup$ – Rodrigo de Azevedo Jun 8 '16 at 12:56
  • 1
    $\begingroup$ Yeah, I'm also curious about the downvote $\endgroup$ – Yair Daon Jun 8 '16 at 13:58
  • 2
    $\begingroup$ @Yair I haven't downvoted it, but it looks like a misleading and incomplete answer. Simply writing a distribution in terms of a Dirac delta doesn't make it continuous! This answer is also limited in that (a) it does not consider general discrete distributions, which may have a countable infinity of atoms and (b) it implicitly assumes $X$ and $Y$ are independent. $\endgroup$ – whuber Jun 8 '16 at 14:07
  • $\begingroup$ @whuber I agree with (b). However, it is said that a discrete RV "can be thought of as...", so I think it adds an interesting view. $\endgroup$ – Yair Daon Jun 8 '16 at 14:15
  • 2
    $\begingroup$ This is why I wrote that your answer is misleading. Because the question concerns the distinction between discrete and continuous distributions--and that distinction is a matter of mathematical definition, not "taste"--your efforts to confuse the two are likely to be less than helpful. $\endgroup$ – whuber Jun 8 '16 at 14:24

This answer assumes that $X$ and $Y$ are independent. Here is a solution which does not need that assumption.

Edit: I am assuming that "continuous" means "having a pdf." If continuous is instead intended to mean atomless, the proof is similar; simply replace "Lebesgue null set" with "singleton set" in what follows.

Let the support of $X$ be the countable set $\{x_1,x_2,x_3\dots\}$. I will use

Lemma: A random variable $Z$ is continuous if and only $P(Z\in E)=0$ for all Borel measurable sets $E$ with Lebesgue measure zero.

Proof: Use the Lebesgue-Radon-Nikodym theorem. $ \square$

To prove $X+Y$ is continuous, take any null set $E$, and note that $$ P(X+Y\in E)=\sum_k P(\{Y+x_k\in E\}\cap \{X=x_k\})\le \sum_k P(Y+x_k\in E) $$ But $Y+x_k\in E$ if and only if $Y\in E-x_k$. The shifted set $E-x_k$ is still Lebesgue null. Since $Y$ is continuous, this means $P(Y+x_k\in E)=0$, so the above summation is zero, proving $X+Y$ is continuous.

For the question of products, the same logic applies as long as $P(X=0)=0$. If $P(X=0)=1$, then $XY$ is discrete with $P(XY=0)=1$. Otherwise, $XY$ is a nontrivial mixture.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.