Find integration on a function of two independent random variables

741 Views Asked by At

Suppose $x,y$ are sampled from a uniform distribution on the interval $[0,1]$.

Find the probability of $z= (2x-1)^2 + (2y-1)^2 \leq 1$, that is, $P(z\leq1)$.

What I've tried:

  1. $P(z\leq 1) = F_z(1)=\int_0^1 z dz$, how can I further simplify it, do I plug $z$ with $x,y$? If so, $z$ is a non-linear function on $x,y$, how would I deal with $dz$ as well as the domains for the integration.

  2. Observing the function form of $z$ is actually an ellipse, can I use the the property of ellipse calculating the area proportion of $z=1$ to the upper bound area $z=2$ so that to obtain the probability $P(z \leq 1)$? Feels like doing so we need some extra assumptions, for example, each ellipse is drawing with the same probability.

1

There are 1 best solutions below

2
On BEST ANSWER

Firstly, let us write capital letters for r.v.'s and small letters for real numbers.

  1. $F_Z(1)\neq \int_0^1 z\,dz = \frac12$.

$$F_Z(1)=\int_0^1 f_Z(z)\,dz, $$ and to use this formula, you need to find probability density function $f_Z(z)$ first, which is much more complicated than solving the original problem.

From the other side, $$ \mathbb P(Z\leq 1)=\mathbb P((2X-1)^2 + (2Y-1)^2 \leq 1)=\iint\limits_{(2x-1)^2 + (2y-1)^2 \leq 1} f_X(x)f_Y(y)\, dx\,dy. $$ Note that $f_X(x)f_Y(y)=1$ for $(x,y)\in[0,1]\times[0,1]$ only. And to find the integral, one need to find the area of intersection of the filled circle $(2x-1)^2 + (2y-1)^2 \leq 1$ and unit square. Find that the circle is inscribed in a square and the desired probability equals to its area ($\pi/4$).

  1. The area $z\leq 2$ is not the area where $(X,Y)$ can be located. As already mentioned, $(X,Y)$ lies in unit square, not in ellipse.