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.

Let $F(x,y)=1$ for $x+y\geq 0$ and be zero otherwise. Show that $F$ cannot possibly be the joint distribution function of a pair of random variables.

Ok so basically I need to show that there can't exist a function $f(x,y)$ such that $\int_{-\infty}^{\infty}\int_{x=-y}^{\infty}f(x,y)dxdy=1$ and that $\int_{-\infty}^{\infty}\int_{-\infty}^{x=-y}f(x,y)dxdy=0$, correct? I haven't explicitly found a counter-example to this (and I suppose if the claim is to be true then there shouldn't be one), but it seems plausible that such an $f(x,y)$ might exist, and playing around with these double integrals hasn't led to any contradictions.

So is there some other requirement on my probability density function $f(x,y)$, or is there some other requirement that the existence of random variables entails? To be honest, I find random variables a bit confusing and their role seems often superfluous and designed to confuse rather than clarify.

Thanks! =)

share|improve this question
    
Your reformulation is not correct. –  GEdgar Jan 2 '13 at 20:47
    
Hint: What is the total probability that a joint distribution assigns to the entire space? What is the 'total probability' of this function? –  Chris Janjigian Jan 2 '13 at 20:55
    
well 1 right??? –  cactuar Jan 2 '13 at 20:56
    
Right. The set where $x + y \geq 0$ is a half plane, so the integral of this function is just the area of half of the plane. What is the area of half of the plane? –  Chris Janjigian Jan 2 '13 at 20:57
    
Well the area of the half plane is infinite.. but I thought we were integrating a probability density function over the half plane, and this function may go to zero quickly in all directions right? –  cactuar Jan 2 '13 at 21:01

1 Answer 1

up vote 7 down vote accepted

So... Assume by way of contradiction that there exists some random variables $X$ and $Y$ such that $F(x,y)=\mathbb P(X\leqslant x,Y\leqslant y)$ for every $x$ and $y$, with $F(x,y)=\mathbf 1_{x+y\geqslant\color{red}{5}}$, say.

Since $\color{green}{6}+\color{blue}{2}\geqslant\color{red}{5}$, $F(\color{green}{6},\color{blue}{2})=F(\color{blue}{2},\color{green}{6})=1$, hence the events $[X\leqslant \color{green}{6},Y\leqslant\color{blue}{2}]$ and $[X\leqslant\color{blue}{2},Y\leqslant \color{green}{6}]$ both have full probability. Their intersection $[X\leqslant\color{blue}{2},Y\leqslant\color{blue}{2}]$ must have full probability as well, that is, $F(\color{blue}{2},\color{blue}{2})=1$. But $\color{blue}{2}+\color{blue}{2}\lt\color{red}{5}$ hence $F(\color{blue}{2},\color{blue}{2})=0$. This is absurd.

share|improve this answer
    
Note that this does not assume that the distribution of $(X,Y)$ has a density, since $F$ could a priori be the PDF of some discrete (or even worse...) distribution on the plane. –  Did Jan 2 '13 at 21:13
    
Oh cool I see it, and the reason this happens in the case of a plane and not on the real line is that the latter is ordered while the former is not! Thanks. –  cactuar Jan 2 '13 at 21:16
    
@did I follow you up till $(-5) + (-5) < 2$. Do you mean $< 0$? –  Calvin Lin Jan 2 '13 at 21:35
    
@CalvinLin No, I meant $\lt2$ because I had used $x+y\geqslant2$ to define $F$, to show there was nothing special with $x+y\geqslant0$. New version uses some "random" parameters, namely $(\color{red}{5},\color{green}{6},\color{blue}{2})$, to make the same idea still more apparent. –  Did Jan 2 '13 at 22:29
    
@did Thanks! The colors helped a lot. I missed the $1_{x+y\geq 5}$ as I thought you were doing the original problem. –  Calvin Lin Jan 2 '13 at 23:01

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.