8

I was reading this question. It is about notation but I would like to ask something regarding the sum of two normally distributed random variables. If $X$ is a normally distributed random variable with mean $\mu_{X}$ and variance $\sigma_{X}^{2}$, and $Y$ is a normally distributed random variable with mean $\mu_{Y}$ and variance $\sigma_{Y}^{2}$, what kind of normal distribution does $X+Y$ follow when

  • $X$ and $Y$ are independent,
  • $X$ and $Y$ are uncorrelated,
  • $X$ are $Y$ are correlated?

How does one give a proof for each case?

dada
  • 97
  • 1
  • 1
  • 2
  • 7
    Hi dada and welcome to the site! Did you look at the [Wikipedia page](https://en.wikipedia.org/wiki/Sum_of_normally_distributed_random_variables) concerning this problem? It contains proof of all listed cases. If yes, in what way did you find it wanting? – COOLSerdash Jul 21 '15 at 11:02

2 Answers2

22

Regardless of whether $X$ and $Y$ are normal or not, it is true (whenever the various expectations exist) that \begin{align} \mu_{X+Y} &= \mu_X + \mu_Y\\ \sigma_{X+Y}^2 &= \sigma_{X}^2 + \sigma_{Y}^2 + 2\operatorname{cov}(X,Y) \end{align} where $\operatorname{cov}(X,Y)=0$ whenever $X$ and $Y$ are independent or uncorrelated. The only issue is whether $X+Y$ is normal or not and the answer to this is that $X+Y$ is normal when $X$ and $Y$ are jointly normal (including, as a special case, when $X$ and $Y$ are independent random variables). To forestall the inevitable follow-up question,

No, $X$ and $Y$ being uncorrelated normal random variables does not suffice to assert normality of $X+Y$. If $X$ and $Y$ are jointly normal, then they also are marginally normal. If they are jointly normal as well as uncorrelated, then they are marginally normal (as stated in the previous sentence) and they are independent as well. But, regardless of whether they are independent or dependent, correlated or uncorrelated, the sum of jointly normal random variables has a normal distribution with mean and variance as given above.


In a comment following this answer, ssdecontrol raised another question: is joint normality just a sufficient condition to assert normality of $X+Y$, or is it necessary as well?

  • Is it possible to find marginally normal $X$ and $Y$ that are not jointly normal but their sum $X+Y$ is normal? This question was asked in the comments below by Moderator Glen_b. This is indeed possible, and I have given an example in an answer to this question.

  • Is it possible to find $X$ and $Y$ such that they are not jointly normal but their sum $X+Y$ is normal? Here, we do not insist on $X$ and $Y$ being marginally normal. The answer is Yes, and an example is given by kjetil b halvorsen. Another, perhaps simpler, answer is as follows. Consider $U$ and $V$ be independent standard normal random variables and $W$ a discrete random variable taking on each of the values $+1$ and $-1$ with probability $\frac 12$. Then, $X = U+W$ and $Y=V-W$ are not marginally normal (they have identical mixture Gaussian density $\frac{\phi(t+1)+\phi(t-1)}{2}$), and so are not jointly normal either. But their sum $X+Y = U+V$ is a $N(0,2)$ random variable.

Dilip Sarwate
  • 41,202
  • 4
  • 94
  • 200
  • +1, especially for the important clarification. Is joint normality necessary, or just sufficient? – shadowtalker Dec 26 '15 at 23:11
  • 1
    @ssdecontrol: It is not necessary. – kjetil b halvorsen Dec 26 '15 at 23:25
  • 2
    @ssdecontrol See [this answer of mine](http://stats.stackexchange.com/a/28808/6633) for a reference to another interesting result: If $X$ and $Y$ are independent, and their sum $X+Y$ is normal, then $X$ and $Y$ must themselves be normal too. – Dilip Sarwate Dec 27 '15 at 00:15
  • 1
    @kjetil do you have an example where you don't have joint normality (joint normality including the cases of independence and perfect correlation), but where the sum is normal and the margins are normal? – Glen_b Dec 27 '15 at 05:48
  • @Glen_b: Only a rather artificial example, like let $X$ be standard normal, $Y$ equal to $X$ if $\lfloor X \rfloor$ is even, else 0, and $Z$ defined by the opposite condition. Then $(Y,Z)$ is not multinormal but their sum is $X$ so normal. Any more natural examples? Here $\lfloor \cdot \rfloor$ is the floor function, the largest integer not greater than the argument. – kjetil b halvorsen Dec 30 '15 at 20:05
  • @Dilip For the one you attribute to me, try this: Let $U,V$ be iid $N(0,1)$. In the first quadrant (i.e. $U>0,V>0$) let $X=\min(U,V)$ and $Y = \max(U,V)$. For the other quadrants, rotate this mapping about the origin, like [this](http://i.stack.imgur.com/Resul.png) (the purple represents regions with doubled probability and the white regions are ones with no probability). Then by symmetry both $X$ and $Y$ are standard normal but are clearly not bivariate normal, and $X+Y = U+V$ which is $\sim N(0,2)$ – Glen_b Jan 04 '16 at 12:04
  • @Glen_b I did not attribute any example to you, only the question regarding marginally normal but not jointly normal random variables whose sum is normal. That being said, I am not sure I know what iid normal random variables confined to the first quadrant mean. Would you care to write up your example more fully as an answer to [this question](http://stats.stackexchange.com/q/189130/6633) which I posted a short while ago? – Dilip Sarwate Jan 04 '16 at 19:34
  • @Dilip By "the one you attributed to me" I mean the question you attributed to me. In my comment above I don't mean "normal variables confined to the first quadrant" -- I transform a pair of iid normals (U,V) -> (X,Y), by explaining how the transformation works in the first quadrant; the transformation in the other quadrants are the same thing, rotated about the origin. I should be able to post an answer in a few hours. – Glen_b Jan 04 '16 at 22:20
0

Let $Z = X + Y$ be, when X and Y are independent you can prove it using the moment-generating function

$ M_{X,Y}(t) = E(\exp(t(X+Y)) \\ M_{X,Y}(t)= E(\exp\{tX\}\exp\{tY\}))\\ M_{X,Y}(t) = M_x(t)M_Y(t) \\ M_{X,Y}(t) = \exp\{ (\mu_X + \mu_Y)t + \frac{\sigma^2_X + \sigma^2_Y}{2}t^2 \} $

Then, $Z \sim N( \mu_X + \mu_Y, \sigma^2_X + \sigma^2_Y )$

Acquijn
  • 1
  • 1