Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I have an exercise where I want to calculate $E[X-Y]$ where $X,Y$ are discrete R.V.

Since the mean is defined with a sum I hope that $E[X-Y]=E[X]-E[Y]$ but I am having a hard time checking it.

$X,Y$ are discrete so $$E[X]=\sum_{i=1}^N x_i p_i,\quad E[Y]=\sum_{j=1}^M y_i q_j$$ where $p_{i},q_{j}>0$ and $\sum p_{i}=\sum q_i = 1$.

So $$E[X]+E[Y]=\sum_{i=1}^N x_i p_i + \sum_{j=1}^M y_i q_j$$ and $$E[X+Y]=\sum_{i,j}(x_i+y_j)p'_{ij}.$$ But I don't know what $p'_{ij}$ is or how to continue.

I would appreciate any help here!

share|improve this question
2  
Much more is true, always $E(\sum a_iX_i)=\sum a_i E(X_i)$. –  André Nicolas Dec 2 '12 at 18:21
    
@AndréNicolas - good to hear that! for some reason I'm having problems with this. I checked this for the continues case and I think that it was trivial since it was the same $x$ that multiplys the density functions, but in this case there are different probabilitys so it doesn't add up like in the continues case –  Belgi Dec 2 '12 at 18:24
    
Maybe try conditional probability –  Lior B-S Dec 2 '12 at 18:25
    
@MichaelHardy - thanks for the edit. are you using lyx ? I wonder how to do sums display like this... –  Belgi Dec 2 '12 at 18:38
    
I've never used LyX and I've just barely ever heard of it. If you write \Sigma_{i=1}^N and \sum_{i=1}^N in a "displayed" setting, they get rendered like this: $\displaystyle\Sigma_{i=1}^N$ and $\displaystyle\sum_{i=1}^N$. In an "inline" setting, they look like this: $\Sigma_{i=1}^N$ and $\sum_{i=1}^N$. Using \sum in this context is standard. –  Michael Hardy Dec 2 '12 at 18:56

1 Answer 1

up vote 3 down vote accepted

I suspect that in your exercise you are supposed to already use, as a known-proven fact, the linearity of expectation: $E[aX+bY]=a E[X] + b E[Y]$, but anyway:

With your notation, $E[X+Y]=\sum_{i,j}(x_i+y_j)p'_{ij} =\sum_{i,j}x_{i}p'_{ij}+\sum_{i,j}y_{j}p'_{ij}$ where $p'_{ij}$ is the probability that $X$ takes the value $x_i$ and $Y$ takes the value $y_j$.

But the first term is $ \sum_i \sum_j x_i p'_{ij}= \sum_i x_i (\sum_j p'_{ij}) = \sum_i x_i p_i = E(X)$, and analogously the second is $E(Y)$.

Updated: $\sum_j p'_{ij}=p_i$ because it's the sum of probabilities for a fixed $x_i$ and for all the possible values of $y_j$; this is known (both for discrete and continuous distributions) as the "marginal distribution". http://en.wikipedia.org/wiki/Marginal_distribution

share|improve this answer
    
I don't understand the equality on the last paragraph, can you please add some more details ? what is $p'_{i,j}$ ? is it $p_{i}p_{j}$ ? –  Belgi Dec 2 '12 at 18:29
    
No, $p'_{i,j}$ is what I defined in the second parag., it's the joint probabilty. You don't know its form, and don't need to. –  leonbloy Dec 2 '12 at 18:33
    
I see, I think I get it. thanks! –  Belgi Dec 2 '12 at 18:35

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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