# Geometric Distribution

1. Apr 5, 2012

### topgun08

Question:
Two faulty machines, M1 and M2, are repeatedly run synchronously in parallel (i.e., both machines execute
one run, then both execute a second run, and so on). On each run, M1 fails with probability p1 and M2 with
probability p2, all failure events being independent. Let the random variables X1, X2 denote the number of
runs until the ﬁrst failure of M1, M2 respectively; thus X1, X2 have geometric distributions with parameters
p1, p2 respectively.
Let X denote the number of runs until the ﬁrst failure of either machine. Show that X also has a geometric
distribution, with parameter p1 + p2 − p1p2

X1 has a geometric distribution of (1-p1)^i-1 * p1
X2 has a geometric distribution of (1-p2)^i-1 * p2

I'm confused an don't know how to proceed. Any help is appreciated.

2. Apr 5, 2012

### chiro

Hey topgun08.

I'll start with one hint: in terms of failures, consider the probability distribution P(X1 OR X2)

3. Apr 5, 2012

### topgun08

After much deliberation here's what I arrived at.

Consider both machines as biased coins such that Tails means the machine fails and Heads means it runs. Thus for Machine1 and Machine 2,
Pr[T1] = p1 and Pr[H1] = 1-p1
Pr[T2] = p2 and Pr[H2] = 1-p2

So the running of both machines can be considered as flipping both these biased coins together, giving us the below table:

Pr[H1H2] = (1-p1)(1-p2)
Pr[H1T2] = (1-p1)(p2) = p2-p1p2
Pr[T1H2] = (p1)(1-p2) = p1-p1p2
Pr[T1T2] = (p1)(p2) = p1p2

And the questions asks for when either machine fails, so adding the probabilities that contain a tails together:
Pr[H1T2] + Pr[T1H2] + Pr[T1T2] = p2-p1p2 + p1-p1p2 + p1p2 = p2 + p1 - p1p2

Which is equal to what the question gave us. Please let me know if my logic is sound or if there's a better way to solve this, and thank you for your help!

4. Apr 5, 2012

### chiro

That looks pretty good, and the logic looks very sound.

Since events are independent you can apply product rule and since the events are disjoint you can simply add them together so under these two assumptions the algebra should work.

If the events were not independent or the events were not disjoint this would not work, but since this is not the case it should be ok.