0

Let $(X,Y)$ be a point chosen at random from the triangle $\{x,y:0\leq x\leq y\leq 1\}$. $f_{X,Y}(x,y)=2$ if $(x,y)$ is in the triangle, and it is 0 otherwise. Find the probability density function for $X$.

What confuses me about this problem is understanding how $x$ and $y$ make up the triangle. If I'm understanding this correctly, then the biggest triangle we can make has vertices $(0,1),(1,1),$ and $(0,0)$. If this is the case then the probability should be 1. As $x$ increases, $y$ can at least be $x$ which means that $y$ is dependent on $x$. From this though I'm not sure where to go from here.

Peetrius
  • 680

1 Answers1

2

If I'm understanding this correctly, then the biggest triangle we can make has vertices $(0,1),(1,1),$ and $(0,0)$.

You're right, that's the support of your random (multivariate) variable.

If this is the case then the probability should be 1.

Huh... what? What you know is that "total" probability is $1$, i.e. $\int_{-\infty}^{\infty} \int_{-\infty}^{\infty} f_{X,Y}(x,y) dx dy=1$. Now, because the density here is constant, denoting by $S$ the support region and by $A_S$ its area we get $\int_{S} 2 dx dy= 2 \, A_S= 1$ and this is indeed true, because the area of the triange is $\frac12$. Then, it's all right.

As $x$ increases, $y$ can at least be $x$ which means that $y$ is dependent on $x$.

Yes. In fact, if you have a bounded support that it's not a rectangle, (or a cartesian product of rectangles) then the variables are dependent.

You have the joint density $f_{X,Y}$. To get the single variable ("marginal") density, you sum (integrate) over the other variable ("marginalize") :

$$ f_X(x)=\int f_{X,Y}(x,y) dy $$

Because the density is constant, and the support is known, what remains is just to get the integrations limits right, i.e. which is the range for the integrating variable ($y$) for each fixed $x$. Can you go on from here?

leonbloy
  • 66,202
  • I gave this a bit more thought. Would we be doing a double integral over the triangle? The first from 0 to $x$ then the second from $x$ to $y$? – Peetrius Apr 17 '19 at 19:15
  • No, the double integral is what gives 1. To marginalize you need only to integrate with respect to $y$, as I explained. – leonbloy Apr 17 '19 at 19:17
  • So is $f_{X,Y}(x,y)$ simply $(1-x)$ which means I take $\int_0^1(-x+1)dy$ to find $f_X(x)$? Sorry, this feels far beyond my skills in probability. I'm going to offer a bounty on this problem. – Peetrius Apr 19 '19 at 00:14
  • Using the last formula $$f_X(x)=\int_x^1 2dy = 2(1-x).$$ –  Apr 19 '19 at 00:28
  • 1
    NB: You should always include the support.$$f_X(x)= 2(1-x)~\mathbf 1_{0\leqslant x\leqslant 1}$$ – Graham Kemp Apr 19 '19 at 02:05
  • Ah, I see. So to derive the cumulative density function for $X$ we would simply integrate respect to $x$ from 0 to $x$ then? – Peetrius Apr 19 '19 at 16:38