HW5 Sols

(c)2017 UM Math Dept
licensed under a Creative Commons

By-NC-SA 4.0 International License.
Homework 5
Math 425: Summer 2017
due Friday, August 4
All solutions to written homework problems must be clearly written, with explanations.
A number or sequence of equations without explanation will not get credit. You are
encouraged to work with other students on the homework, but solutions must be written
independently. Homework is due at the start of class (1 pm) on the due date.
1. Suppose that an appliance is constructed in such a way that it requires that n inde-
pendent electronic components are all functioning. Assume that the lifespan of each
of these, Tj , is an exponential random variable with parameter j .
(a) Let X be the random variable giving the lifespan of the appliance. Find the CDF
and PDF for X.
(b) Find the expected value of X. Then find the median lifespan of the appliance
(that is, the time t at which half of the appliances are likely to have broken
and half to be working). Which will the manufacturer of the appliance use in
advertising?
Solution: (a) For the appliance to work we require that all of the n components be
functioning. Thus
P {X t} = P {T1 t T2 t Tn t}
= P {T1 t} P {T2 t} P {Tn t}
= e1 t e2 t en t = e(1 +2 ++n )t .
j , the CDF and PDF of X are F (t) = 1 ekt

P P
Thus X Exp( j ). Letting k =
and f (t) = kekt , both for t 0.
(b) We know that E[X] = k1 . The median is the t for which P {X t} = P {X t} =
1
2
, or ekt = 21 . Thus the median is m = k1 ln(2) 0.693 k1 = 0.693E[X]. Thus the
manufacturer will report the expected value (average) rather than the median because
it indicates a longer lifetime.
2. Suppose Yj (1 j n) are independent exponential random variables with pa-

rameters j . Find the distribution and density function for the random variable
M = min(Y1 , Y2 , Y3 , . . . , Yn ). What type of random variable is M ?
Hint: The cdf FM (m) = P {M m} = 1 P (?).
Solution: The cumulative distribution function, FM (m) = P {M m}. This is
FM (m) = 1 P {all Yj > m}. Note that this is the same situation as in problem 1!
The CDF is therefore
FM (m) = 1 P {all Yj > m}
= 1 P (Y1 > m)P (Y2 > m) P (Yn > m)
= 1 e(1 ++n )m .
Thus M Exp(1 + + n ), and exponential random variable, and

fZ (z) = (1 + + n )e(1 ++n )z .
3. Show that geometric random variables are memoryless. Explain why this makes sense.
Interesting fact: geometric random variables are the only discrete, memoryless random
variables, and exponentials are the only continuous ones.
Solution: Given X geometric(p), we need to show that P (X > s + t|X > t) =
P (X > s). Now P (X > s) = (1 p)s , since the event that we need more than s
trials for a success is the same as the event that the first s trials are failures. (Or, you
can show this mathematically by computing the relevant summation; appropriately
enough, you will need to use the formula for the sum of a geometric series.) Next,
P (X > s + t X > t) P (X > s + t) (1 p)s+t
P (X > s+t|X > t) = = = = (1p)s .
P (X > t) P (X > t) (1 p)t
Or, one can argue as follows. A geometric r.v. X with parameter p gives the number
of independent trials required for the first success. Thus the probability we need more
than k trials is (1 p)k as above. Now suppose we have already had m failures, so
we know X > m. Whats the probability we need more than k additional trials? The
probabilities of success dont change because the trials are independent; its as if we
are beginning all over again, forgetting our m failures, and so the probability we will
need more than m + k trials in total is also (1 p)k .
4. Let X exponential(), and let Y = eX .

(a) Find the cdf of Y .
(b) Use (a) to find the pdf of Y . Check that your answer really is a density by showing
that it integrates to 1.
(c) Find the pdf of Y again, this time using shortcut formula (Theorem 7.1 p. 209),
being sure to explain why it applies.
Solution: (a) Note that the possible values of Y are [1, ). For y in this interval,
FY (y) = P (Y y) = P (eX y)
= P (X ln(y))
= FX (ln(y))
= 1 e ln(y)
One can simply this formula to 1 y1 properties of e and ln. For (b), we simply
differentiate to obtain fY (y) = y e ln(y) or y+1

.
(c) We can use this formula because g(x) = ex is increasing and differentiable. Note
g 1 (y) = ln(y) and fX (x) = ex . Then, as above,
1
fY (y) = fX (g 1 (y))|(g 1 (y))0 | = e ln(y) .
y
5. A bin of five transistors is known to contain two that are defective. They are to be
tested, one at a time, until the defective ones are identified. Let X be the number of
tests needed to find one defective transistor, and let Y be the additional tests needed
to find the second.
(a) Find the joint pmf of X and Y .

(b) Find the probability that all five transistors end up being tested.
(c) Are X and Y independent? Justify.
Solution: Note: there are two ways to interpret this problem; both are discussed here.
For example, the probability that X = 1 and Y = 1 is 25 14 = 10 1
, the probability we
choose the two bad transistors in a row (without replacement, since with replacement
would be a silly way to conduct the test). Similarly, the probability that X = 1 and
1
that we then draw good, good, and bad transistors is also 10 . Note Y = 3 in this case.
Now the probability that X = 1 and that we then draw all three good transistors is
1
also 10 ; but in this case theres only one transistor left in the bin, and since we know
there are two bad transistors, we dont need to test the last one to identify it, so Y = 3
here too. If you thought this way, your pmf should look like:
Similarly to the argument above, X = 3 might mean drawing good, good, bad, and so
we need one more test to determine the last defective one, or good, good, good in
which case we dont need any more tests and Y = 0.
If you assumed that we would actually test the transistors to make sure they were bad,
your pmf should look like this instead:
(b) For the first pmf, it is 0; we never need more than 4 tests. For the second pmf, it
4
is 10 , the sum of the probabilities along the backward diagonal.
(c) No. You can test that the joint pmf is not the product of the marginals, or note
that knowing X often restricts the possibilities for Y , or vice versa.
6. Suppose X and Y have the joint density function fX,Y (x, y) = e(x+y) for x 0, y 0.
Let W = X + Y .
(a) Are X and Y independent? Justify.
(b) Find the cdf of W .
(c) Use (b) to find the pdf of W .
(d) Now find the pdf of W using the convolution formula (equation (3.2) on p. 240).
Solution:
(a) Yes. You can compute the marginals ex and ey and note that their product is
the joint pdf, or you can just note that we are able to factor the joint pdf into x and
y pieces and there are separate x and y conditions on the domain.
(b) We want FW (w) = P (W w) = P (X + Y w) = P (Y X + w). Now, our
sample space is the whole first quadrant, so we need to integrate over the region in this
quadrant under the line y = x + w (draw the picture!). Note that if w < 0 there is
no such region, so FW (w) = 0 there. We thus find that
Z w Z wx Z w Z wy
(x+y)
FW (w) = e dydx or e(x+y) dxdy
0 0 0 0
w w
=1e we , for w [0, ).
(c) Differentiating, we obtain fW (w) = wew .

(d) Using the convolution formula for the density of a sum of independent r.v.s, we
obtain
Z
fX+Y (a) = fX (a y)fY (y)dy

Z a
= e(ay) ey dy = aea
0
as above (note a and w are just variable names; its the same function). The difficult
part here is figuring out the limits 0 and a. Recall that the marginals for X and Y
are ex and ey , but take care: this is only true for x and y greater than or equal to
0. Thus the integrand above is only nonzero when we have both a y 0, in other
words y a, and y 0.

HW5 Sols

Uploaded by

Copyright:

Available Formats

HW5 Sols

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

HW5 Sols

Uploaded by

Copyright:

Available Formats

(c)2017 UM Math Dept

licensed under a Creative Commons

j , the CDF and PDF of X are F (t) = 1 ekt

2. Suppose Yj (1 j n) are independent exponential random variables with pa-

Thus M Exp(1 + + n ), and exponential random variable, and

4. Let X exponential(), and let Y = eX .

(a) Find the joint pmf of X and Y .

(c) Differentiating, we obtain fW (w) = wew .

You might also like