Massachusetts Institute of Technology

Department of Electrical Engineering & Computer Science

6.041/6.431: Probabilistic Systems Analysis
(Fall 2010)

Problem Set 6: Solutions

1. Let us draw the region where fX,Y (x, y) is nonzero:

y

y-x=z

0 1 2 x

� x=2 � y=x
The joint PDF has to integrate to 1. From x=1 y=0 ax dy dx = 73 a = 1, we get a = 37 .
 �
2 3  9

 1 7 x dx, if 0 ≤ y ≤ 1,  14 , if 0 ≤ y ≤ 1,
� �2 3 3 2
(b
a)) fY (y) = fX,Y (x, y) dy = y 7 x dx, if 1 < y ≤ 2,
= (4 − y ), if 1 < y ≤ 2,
 14
0, otherwise.

 0, otherwise
(c)

3 fX,Y (x, 32 ) 8 3
fX|Y (x | ) = 3 = x, for 2 ≤ x ≤ 2 and 0 otherwise.
2 fY ( 2 ) 7

Then, � � 2
1 3 18 4
�
E |Y = = x dx = .
X 2 3/2 x7 7

(d) We use the technique of first finding the CDF and differentiating it to get the PDF.

FZ (z) = P(Z ≤ z)
= P(Y − X ≤ z)



 0, if z < −2,

 � x=2 � y=x+z
 3
x dy dx = 87 + 67 z − 1 3

14 z , if −2 ≤ z ≤ −1,


7

x=−z y=0
=
 x=2 y=x+z 3

� �
9


 x dy dx = 1 + 14 z, if −1 < z ≤ 0,



 x=1 y=0 7
 1, if 0 < z.

6 3 2

d  7 − 14 z , if −2 ≤ z ≤ −1,
9
fZ (z) = FZ (z) = 14 , if −1 < z ≤ 0,
dz
0, otherwise.


Page 1 of 4
 Massachusetts Institute of Technology
Department of Electrical Engineering & Computer Science
6.041/6.431: Probabilistic Systems Analysis
(Fall 2010)

2. The PDF of Z, fZ (z), can be readily computed using the convolution integral:
� ∞
fZ (z) = fX (t)fY (z − t) dt.
−∞

For z ∈ [−1, 0],

z
z3 2
� �
1 3 1
�
fZ (z) = · (1 − t2 ) dt = z− + .
−1 3 4 4 3 3
For z ∈ [0, 1],
z
z 3 (z − 1)3
� �
1 3 1
�
fZ (z) = · (1 − t2 ) dt = 1− + .
z−1 3 4 4 3 3
For z ∈ [1, 2],
1 z−2
(z − 1)3 2(z − 2)3
� �
1 3 2 3 1
� �
fZ (z) = · (1 − t2 ) dt + · (1 − t2 ) dt = z+ − −1 .
z−1 3 4 −1 3 4 4 3 3
For z ∈ [2, 3],
z−2
2 3 1 �
�
· (1 − t2 ) dt = 3 + (z − 3)3 − (z − 2)3 .
�
fZ (z) =
z−3 3 4 6
For z ∈ [3, 4],
1
2 3 1 �
�
· (1 − t2 ) dt = 11 − 3z + (z − 3)3 .
�
fZ (z) =
z−3 3 4 6
A sketch of fZ (z) is provided below.

1
fZ (z)

0.5

−1 0 1 2 3 4
z
3. (a) X1 and X2 are negatively correlated. Intuitively, a large number of tosses that result in a 1
suggests a smaller number of tosses that result in a 2.
(b) Let At (respectively, Bt ) be a Bernoulli random variable that is equal to 1 if and only if the
tth toss resulted in 1 (respectively, 2). We have E[At Bt ] = 0 (since At =

6 0 implies Bt = 0)
and

1 1
E[At Bs ] = E[At ]E[Bs ] = · for s =6 t.
k k
Thus,

E[X1 X2 ] = E [(A1 + · · · + An )(B1 + · · · + Bn )]

1 1
= nE [A1 (B1 + · · · + Bn )] = n(n − 1) · ·
k k

Page 2 of 4
 Massachusetts Institute of Technology
Department of Electrical Engineering & Computer Science
6.041/6.431: Probabilistic Systems Analysis
(Fall 2010)

and
cov(X1 , X2 ) = E[X1 X2 ] − E[X1 ]E[X2 ]
n(n − 1) n2 n
= 2
− 2 = − 2.
k k k
The covariance of X1 and X2 is negative as expected.
4. (a) If X takes a value x between −1 and 1, the conditional PDF of Y is uniform between −2
and 2. If X takes a value x between 1 and 2, the conditional PDF of Y is uniform between
−1 and 1.
Similarly, if Y takes a value y between −1 and 1, the conditional PDF of X is uniform
between −1 and 2. If Y takes a value y between 1 and 2, or between −2 and −1, the
conditional PDF of X is uniform between −1 and 1.
(b) We have 
 0, if −2 ≤ y ≤ −1,
E[X | Y = y] = 1/2, if −1 < y ≤ 1,
0, if 1 ≤ y ≤ 2,


and 
 1/3, if −2 ≤ y ≤ −1,
var(X | Y = y) = 3/4, if −1 < y ≤ 1,
1/3, if 1 ≤ y ≤ 2.


It follows that E[X] = 3/10 and var(X) = 193/300.

(c) By symmetry, we have E[Y | X] = 0 and E[Y ] = 0. Furthermore, var(Y | X = x) is the
variance of a uniform PDF (whose range depends on x), and
�
4/3, if −1 ≤ x ≤ 1,
var(Y | X = x) =
1/3, if 1 < x ≤ 2.
Using the law of total variance, we obtain
4 4 1 1
var(Y ) = E[var(Y | X)] = · + · = 17/15.
5 3 5 3
5. First let us write out the properties of all of our random variables. Let us also define K to be the
number of members attending a meeting and B to be the Bernoulli random variable describing
whether or not a member attends a meeting.
1 p
E[N ] = , var(N ) = ,
1−p (1 − p)2
1 1
E[M ] = , var(M ) = 2 ,
λ λ
E[B] = q, var(B) = q(1 − q).

(a) Since K = B1 + B2 + · · · BN ,
q
E[K] = E[N ] · E[B] = ,
1−p
q(1 − q) pq 2
var(K) = E[N ] · var(B) + (E(B))2 · var(N ) = + .
1−p (1 − p)2

Page 3 of 4
 Massachusetts Institute of Technology
Department of Electrical Engineering & Computer Science
6.041/6.431: Probabilistic Systems Analysis
(Fall 2010)

(b) Let G be the total money brought to the meeting. Then G = M1 + M2 + · · · + MK .

q
E[G] = E[M ] · E[K] = ,
λ(1 − p)
var(G) = var(M ) · E[K] + (E[M ])2 var(K)
pq 2
� �
q 1 q(1 − q)
= + + .
λ2 (1 − p) λ2 1−p (1 − p)2

G1† . (a) Let X1 , X2 , . . . Xn be independent, identically distributed (IID) random variables. We note
that
E[X1 + · · · + Xn | X1 + · · · + Xn = x0 ] = x0 .
It follows from the linearity of expectations that

x0 = E[X1 + · · · + Xn | X1 + · · · + Xn = x0 ]
= E[X1 | X1 + · · · + Xn = x0 ] + · · · + E[Xn | X1 + · · · + Xn = x0 ]

Because the Xi ’s are identically distributed, we have the following relationship.

E[Xi | X1 + · · · + Xn = x0 ] = E[Xj | X1 + · · · + Xn = x0 ], for any 1 ≤ i ≤ n, 1 ≤ j ≤ n.

Therefore,

nE[X1 | X1 + · · · + Xn = x0 ] = x0
x0
E[X1 | X1 + · · · + Xn = x0 ] = .
n

(b) Note that we can rewrite E[X1 | Sn = sn , Sn+1 = sn+1 , . . . , S2n = s2n ] as follows:

E[X1 | Sn = sn , Sn+1 = sn+1 , . . . , S2n = s2n ]

= E[X1 | Sn = sn , Xn+1 = sn+1 − sn , Xn+2 = sn+2 − sn+1 , . . . , X2n = s2n − s2n−1 ]
= E[X1 | Sn = sn ],

where the last equality holds due to the fact that the Xi ’s are independent. We also note
that
E[X1 + · · · + Xn | Sn = sn ] = E[Sn | Sn = sn ] = sn .
It follows from the linearity of expectations that

E[X1 + · · · + Xn | Sn = sn ] = E[X1 | Sn = sn ] + · · · + E[Xn | Sn = sn ].

Because the Xi ’s are identically distributed, we have the following relationship:

E[Xi | Sn = sn ] = E[Xj | Sn = sn ], for any 1 ≤ i ≤ n, 1 ≤ j ≤ n.

Therefore,
sn
E[X1 + · · · + Xn | Sn = sn ] = nE[X1 | Sn = sn ] = sn ⇒ E[X1 | Sn = sn ] = .
n

† Required for 6.431; optional for 6.041 Page 4 of 4

MIT OpenCourseWare
http://ocw.mit.edu

6.041 / 6.431 Probabilistic Systems Analysis and Applied Probability

Fall 2010

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.