Let (X,Y) be uniformly distributed over $B=\{(x,y) \in \mathbb{R}^2: x^2+y^2 \leq 1 \}$ resp. $Q=[-1,1]^2$. Now I want to calculate the conditional distributions and of Y given X=x. And then the same for $R=\sqrt(X^2+Y^2)$ and $\varphi = arctan(Y/X)$ (conditional distributions of $\varphi$ given R=r).

At first I calculated the densities: For the unit circle I get: f(x,y) = $\begin{cases} 1/\pi~~~ if~ x^2+y^2 \leq 1\\ 0 ~~~~~~ else \end{cases}$. $f_Y(y) = \frac{2 \sqrt(1-y^2)}{\pi}$

$f_X(x)= \frac{2 \sqrt(1-x^2)}{\pi}$

For the square I get: $f(x,y)= \begin{cases} 1/4~~~if (x,y) \in Q\\ 0~~~~~~else \end{cases}$.


How do I get now the conditional distributions? And how do I get the conditional distributions of $\varphi$?


For the disc, $B$,

$$f_{X|Y}(x|y) = \dfrac{f_{X,Y}(x,y)}{f_Y(y)} = \dfrac{1/\pi}{2\sqrt{1-y^2}/\pi} = \dfrac{1}{2\sqrt{1-y^2}}\qquad\text{for $|x|\leq \sqrt{1-y^2},$ otherwise $0$.}$$


$$f_{Y|X}(y|x) = \dfrac{1}{2\sqrt{1-x^2}}\qquad\text{for $|y|\leq \sqrt{1-x^2},$ otherwise $0$.}$$

So both these conditional distributions are uniform, which is as we would expect.

For $R,\varphi,\;$ we can use the "change of variable" method. Since $R=\sqrt{X^2+Y^2}$ and $\varphi=\arctan{\left(Y/X\right)}$, we have

$$X = R\cos\varphi \\ Y = R\sin\varphi.$$

So our Jacobian is

$$J=\begin{vmatrix} \dfrac{\partial{x}}{\partial{r}} & \dfrac{\partial{y}}{\partial{r}} \\ \dfrac{\partial{x}}{\partial{\varphi}} & \dfrac{\partial{y}}{\partial{\varphi}} \\ \end{vmatrix} =\begin{vmatrix} \cos\varphi & \sin\varphi \\ -r\sin\varphi & r\cos\varphi \\ \end{vmatrix} = r(\cos^2 \varphi + \sin^2 \varphi ) = r.$$

$$f_{R,\varphi}(r,\varphi) = f_{X,Y}(x(r,\varphi),\;y(r,\varphi))\; |J| = \dfrac{r}{\pi}\qquad\text{for $0\leq r \leq 1$ and $0\leq \varphi \leq 2\pi$}$$

$$f_R(r) = \int_0^{2\pi}{\dfrac{r}{\pi}\;d\varphi} = \left[ \dfrac{r\varphi}{\pi} \right]_0^{2\pi} = 2r.$$

$$f_{\varphi}(\varphi) = \int_0^1{\dfrac{r}{\pi}\;dr} = \left[ \dfrac{r^2}{2\pi} \right]_0^1 = \dfrac{1}{2\pi}.$$

Note that $f_{R,\varphi}(r,\varphi) = f_R(r)f_{\varphi}(\varphi)$ so $R,\varphi$ are independent. Therefore, $f_{\varphi|R}(\varphi|r)=f_{\varphi}(\varphi)$ and $f_{R|\varphi}(r|\varphi) = f_{R}(r)$.


For the square, $Q$, we use a similar method.

$$f_{X|Y}(x|y) = \dfrac{f_{X,Y}(x,y)}{f_Y(y)} = \dfrac{1/4}{1/2} = \dfrac{1}{2}\qquad\text{for $|x|\leq 1,$ otherwise $0$.}$$


$$f_{Y|X}(y|x) = \dfrac{1}{2}\qquad\text{for $|y|\leq 1,$ otherwise $0$.}$$

This means $X,Y$ are independent.

For $R,\varphi$, we have the same Jacobian, $J$, as above. However, the ranges of the r.v.'s are different and more difficult. Firstly, the joint distribution:

$$f_{R,\varphi}(r,\varphi) = f_{X,Y}(x(r,\varphi),\;y(r,\varphi))\; |J| = \dfrac{r}{4}\qquad\text{for $0\leq \varphi \leq 2\pi$ and} \\ 0\leq R \leq \begin{cases} \dfrac{1}{\cos{\varphi}}, & \text{if $k\pi - \dfrac{\pi}{4}\leq \varphi\leq k\pi + \dfrac{\pi}{4},\quad k\in\mathbb{Z}$} \\ \dfrac{1}{\sin{\varphi}}, & \text{if $k\pi + \dfrac{\pi}{4}\leq \varphi\leq k\pi + \dfrac{3\pi}{4},\quad k\in\mathbb{Z}.$} \\ \end{cases} $$

If $0\leq r\leq 1,\quad f_R(r) = \int_0^{2\pi}{\frac{r}{4}\;d\varphi} = \left[\dfrac{r\varphi}{4}\right]_0^{2\pi} = \dfrac{\pi r}{2}.$

If $1\leq r\leq \sqrt{2},\quad f_R(r) = 8\int_{\arccos{(1/r)}}^{\pi/4}{\frac{r}{4}\;d\varphi} = \left[2r\varphi\right]_{\arccos{(1/r)}}^{\pi/4} = \dfrac{\pi r}{2} - 2r\arccos{(1/r)}.$

If $0\leq r\leq 1$,

$$f_{\varphi|r}(\varphi,r) = \dfrac{f_{R,\varphi}(r,\varphi)}{f_R(r)} = \dfrac{r/4}{\pi r/2} = \dfrac{1}{2\pi}.$$

If $1\leq r\leq \sqrt{2}$,

$$f_{\varphi|r}(\varphi,r) = \dfrac{r/4}{\pi r/2 - 2r\arccos(1/r)} = \dfrac{1}{2\pi - 8\arccos(1/r)}.$$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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