2
$\begingroup$

If I generate a uniform distribution of $X$ ranging from $0$ to $2\pi$ (so $X\sim U(0, 2\pi)$), then the probability distribution of $1-\cos(X)$ appears to be this function:

enter image description here

Is this an analytical function that I can work with directly? And if so, how do I prove this?

$\endgroup$
4
  • 1
    $\begingroup$ Based on inspection and some clues in the text, I can guess that it's an arcsine distribution en.wikipedia.org/wiki/Arcsine_distribution but you'll need to provide more details to know for sure. How is $X$ distributed? If an "even distribution" means "uniform" then there are certain transformations of uniform distributions that are definitely arcsine distributions. $\endgroup$
    – Sycorax
    Apr 24, 2023 at 17:59
  • $\begingroup$ I just simulated from $N(0, 1)$ and then applied your $1-\cos(x)$ transformation, leading to a somewhat different plot. $\endgroup$
    – Dave
    Apr 24, 2023 at 18:01
  • 1
    $\begingroup$ @dave could you try it with a uniform distribution ranging from (0, 2pi)? $\endgroup$ Apr 24, 2023 at 18:53
  • 1
    $\begingroup$ You can find extensive explanations by searching our site. $\endgroup$
    – whuber
    Apr 24, 2023 at 20:11

2 Answers 2

5
$\begingroup$

We can evaluate the distribution function $F(y)$ of $Y := 1 - \cos(X)$ directly as follows.

To begin with, note that as $X$ ranges over $[0, 2\pi]$, the range of $Y$ is from $0$ (achieved at $X = 0$ or $2\pi$) to $2$ (achieved at $\pi$). Therefore, for $y < 0$, $F(y) = 0$ and for $y \geq 2$, $F(y) = 1$.

For $y \in [0, 2)$ (it is helpful to draw the graph of $x \mapsto \cos(x)$ on $[0, 2\pi]$ to determine the region $\{x \in [0, 2\pi]: \cos(x) \geq 1 - y\}$. Also keep in mind that the domain of $x \mapsto \arccos(x)$ is $[-1, 1]$ with range $[0, \pi]$, so mirroring is needed for angle that is greater than $\pi$): \begin{align} & F(y) = P[1 - \cos(X) \leq y] = P[\cos(X) \geq 1 - y] \\ =& P[X \in [0, \arccos(1 - y)] \cup [2\pi - \arccos(1 - y), 2\pi]] \\ =& \frac{\arccos(1 - y)}{\pi}. \end{align}

To summarize, the distribution of $Y$ is given by \begin{align} F(y) = \begin{cases} 0 & y < 0, \\ \frac{\arccos(1 - y)}{\pi} & 0 \leq y < 2, \\ 1 & y \geq 2. \end{cases} \end{align}

Taking derivative of $F$ yields the pdf of $Y$: \begin{align} f(y) = \begin{cases} \frac{1}{\pi\sqrt{1 - (1 - y)^2}} & 0 < y < 2, \\ 0 & \text{ otherwise.} \end{cases} \end{align}

A graph of $f$ looks as follows, which matches the histogram you simulated. As @Sycorax pointed out in the comment, this is Arcsine distribution with support $(0, 2)$.

enter image description here

$\endgroup$
2
$\begingroup$

You can get the result with the delta method: $p(z)=\int_R p(x) \delta(z-(1-cos(x)))dx$. https://en.m.wikibooks.org/wiki/Probability/Transformation_of_Probability_Densities

$\endgroup$
0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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