Does anyone know how to find this pdf? I understand that $[X]\sim\mathrm{ Ge} \left(1-e^{-\frac{1}{a}}\right)$, but I got stuck after that ($[X]$ is the integer part of $X$). I would be really grateful for some input.


Because this looks like a textbook problem, I will sketch an answer and leave some details to be completed by you.

Let the PDF of $X$, which is supported on the positive numbers $(0,\infty)$, be given by the function $f_X$. Write $Y=X-[X]$. Because this value obviously must lie between $0$ and $1$, consider an arbitrary number $y$ with $0\le y \lt 1$. The PDF of $Y$--call it $f_Y$--evaluated at $y$ is the probability that $Y \in [y, y+dy)$ for an infinitesimal $dy$. But this event describes all positive numbers whose fractional part is close to $y$; specifically,

$$X \in [y, y+dy) \cup [y+1, y+1+dy) \cup \cdots \cup [y+n, y+n+dy) \cup \cdots.$$

Because this is a countable union of disjoint events, probabilities for $X$ are given by the density $f_X$ by summing the probabilities of the events:

$$f_Y(y)dy=\Pr(Y\in[y, y+dy)) = \sum_{n=0}^\infty \Pr(X\in[y+n, y+n+dy)) = \sum_{n=0}^\infty f_X(y+n)dy.$$


The lefthand graphic plots the PDF of $X$. Colors distinguish the integer parts of $X$: blue for $[X]=0$, red for $[X]=1$, gold for $[X]=2$, and so on. The formula for $Y$ shifts each colored section leftwards against the vertical axis. The PDF for $Y$ is obtained by stacking the heights of each shifted piece, as shown in the righthand graphic.

Since $f_X(x) = e^{-x/a}$ for any positive $x$, this sum becomes

$$f_Y(y)dy = \sum_{n=0}^\infty \exp(-(y+n)/a)dy = \exp(-y/a)\left(\sum_{n=0}^\infty \exp(-a)^n\right)dy.$$

The sum is a geometric series, easily computed in closed form. Comparing the coefficients of $dy$ on each side yields a nice formula for $f_Y(y)$.


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.