9
$\begingroup$

Let $z = xy$ be a product of two uniform random variables, with $x$ having the range $[a, b)$ and $y$ the range $[c, d)$.

What is the probability density function of $z$, and how is it calculated?

$\endgroup$
9
$\begingroup$

There are different solutions depending on whether $a > 0$ or $a < 0$, $c > 0$ or $c < 0$ etc. If you can tie it down a bit more, I'd be happy to compute a special case for you.

In the case of: $a > 0$ and $c > 0$, there are three special sub-cases:

  • Case 1: $a d > b c$

  • Case 2: $a d < b c$

  • Case 3: $a d = b c$

Here is the output for Case 1 using the mathStatica/Mathematica combo ...

Case 1: $a > 0$, $c > 0$, $a d > b c$

Given $X \sim \text{Uniform}(a,b)$ with pdf $f(x)$, and $Y \sim \text{Uniform}(c,d)$ with pdf $g(y)$ and parameters $a>0$ and $c>0$, we have:

f = 1/(b-a);    domain[f] = {x, a, b} && {0 < a < b, a d > b c};
g = 1/(d-c);    domain[g] = {y, c, d} && {0 < c < d};

Find pdf of the product $Z = X * Y$, say $h(z)$:

h = TransformProduct[{f, g}, z]

enter image description here

with domain of support (just define it on the real line):

domain[h] = {z, -∞, ∞};

Here is a plot of the pdf for Case 1:

PlotDensity[h /. {a -> 2, b -> 4, c -> 1/2, d -> {6, 7, 8}}]

enter image description here

update ...

Case 3: $a > 0$, $c > 0$, $a d == b c$

Case 3 is really just a special case of case 1 ...

Given $X \sim \text{Uniform}(a,b)$ with pdf $f(x)$, and $Y \sim \text{Uniform}(c,d)$ with pdf $g(y)$, and parameters $a>0$ and $c>0$, we have:

f = 1/(b-a);    domain[f] = {x, a, b} && {0 < a < b, a d == b c};
g = 1/(d-c);    domain[g] = {y, c, d} && {0 < c < d};

Find pdf of the product $Z = X * Y$, say $h(z)$:

h = TransformProduct[{f, g}, z]

enter image description here

with domain of support (just define it on the real line):

domain[h] = {z, -∞, ∞};

Here is a plot of the pdf for Case 3, for your set-up, with $a = c$, and $b = d$:

PlotDensity[h /. {a -> Range[5], b -> 12, c -> Range[5], d -> 12}]

enter image description here

... and you are all done :) I might update the mathStatica code to automatically output all the special cases ... for the next version :)

$\endgroup$
  • $\begingroup$ In case 2, of course, we can simply make a change of variables and use case one, as they're symmetric. At the moment, I have two specific cases, both with a>0, c>0, one of which falls into case 1, but the other falls into case 3, as a=c, b=d. I'd also really like to have the probability density function Pz(a,b,c,d) if possible, as I don't have access to Mathematica, and there are a lot of different ranges of z that are significant for this problem, more than I could ask someone to solve for me. $\endgroup$ – Theodore Murdock Apr 29 '13 at 14:54
  • $\begingroup$ @TheodoreMurdock Hi Theo - yes, case 2 by symmetry, and case 3 can be obtained as a special case of case 1 or 2. I've updated the above to include case 3 + plot. $\endgroup$ – wolfies Apr 29 '13 at 15:39
  • $\begingroup$ Oh...sorry for being slow to get it, I just realized your function h is the probability density I need, taking different forms in different regions of the graph. Thanks for the help! Now I just need to plug in the numbers, integrate over the proper interval, and I'll be able to clear up the confusion I started a few years ago over how many dwarves are fat enough and tall enough to wield a two-handed sword in dwarf fortress. $\endgroup$ – Theodore Murdock Apr 29 '13 at 15:52
  • $\begingroup$ In an example to a recent answer, I computed the density of the product of two uniform variables on $[0,1]$. I am pleased to see that my result is confirmed here. (+1) $\endgroup$ – robjohn Mar 28 '14 at 10:39
5
$\begingroup$

The general formula for the pdf of the product $Z=XY$ of two random variables $X$ and $Y$ is given by

$$p_{Z}(z)= \int_{-\infty}^{\infty}\frac{1}{|x|}p_{XY}\left (x,\frac{z}{x}\right )\;dx$$

where $p_{XY}(x,y)$ is the joint distribution of the two random variables. If they are independent you can use $p_{XY}(x,y)=p_X(x)p_Y(y)$.

$\endgroup$

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.