EECE7204

Appl. Prob. & Stoch. Proc.

Prof. M. Stojanovic

Solution to Problem Set 2

Solution 1 (2.7).
Here we do calculations with the Normal (Gaussian) random variable of mean 0 and given variance σ 2 . In notation we
often indicate this as X ∼ N (0, σ 2 ). In order to calculate the probabilities P{|X| ≥ kσ} for integer values k = 1, 2, . . .,
R x normal curve that is distributed as N (0, 1). In particular the so-called error
we need to convert this to the standard
function is defined as erf(x) = √12π 0 exp(− 12 ξ 2 )dξ, for x ≥ 0, and so only includes the right hand side of the N (0, 1)
distribution. Expanding P{|X| ≥ kσ} for k positive, we get P{|X| ≥ kσ} = P{{X ≤ −kσ} ∪ {X ≥ kσ}} which is
somewhat cumbersome, so instead we consider the complementary event P{|X| ≤ kσ} which satisfies P{|X| ≥ kσ} =
1 − P{|X| ≤ kσ}. For this complementary event and k > 0, we have
Z kσ Z kσ Z k
1 ξ2 (a) 2 ξ2 (b) 2 y2
P{|X| ≤ kσ} = √ exp(− 2 )dξ = √ exp(− 2 )dξ = √ exp(− )dy = 2erf(k)
2πσ −kσ 2σ 2πσ 0 2σ 2π 0 2
where (a) is based on the fact that the pdf of N (0, σ 2 ) is symmetric around its mean and (b) is obtained by making
the change of variable y = σξ . Hence, P{|X| ≥ kσ} = 1 − P{|X| ≤ kσ} = 1 − 2erf(k). Alternatively, we can use Q
R +∞ 2
function, defined as Q(x) = √12π x exp(− ξ2 )dξ = 21 − erf(x), to obtain P{|X| ≥ kσ} = 2Q(k).

10 0

10 -5

10 -10

10 -15

10 -20

10 -25
1 2 3 4 5 6 7 8 9 10

Figure P1 (2.7)

Solution 2 (2.17).
Let the number of bulbs produced by A and B be nA and nB , respectively, and n = nA + nB be the total number of the
x x
bulbs. So PA = nnA = 41 and PB = nnB = 34 . Since we have FX (x|A) = (1 − e− 5 )u(x) and FX (x|B) = (1 − e− 2 )u(x),
x x
then FX (x) = FX (x|A)PA + FX (x|B)PB = 14 (1 − e− 5 )u(x) + 43 (1 − e− 2 )u(x). Hence, P{bulb burns at least x
months} = 1 − FX (x). Then, P{bulb burns at least 2 months} = 0.44, P{bulb burns at least 5 months} = 0.15, and
P{bulb burns at least 7 months} = 0.08.
Solution 3 (2.18).
By definition FX (x|A) = P{X ≤ x|X ∈ A}. Using Bayes’ rule, we obtain

P{{X ≤ x} ∩ {X ∈ A}} P{{X ≤ x} ∩ {b < X ≤ a}}  0 x<b
FX (x)−FX (b)
FX (x|A) = = = FX (a)−FX (b) b<x<a
P{X ∈ A} P{b < X ≤ a}
x≥a

1
Solution 4 (2.22).
By definition random variables X and Y are independent if and only if fXY (x, y) = fX (x)fY (y) where the marginal

1
 R +∞ R +∞
pdfs are given by fX (x) = −∞ fXY (x, y)dy and similarly fY (y) = −∞ fXY (x, y)dx. Given the joint pdf, the marginal
pdfs are then obtained as
Z +∞ √ Z +∞
1 − 12 [( x3 )2 + y2 )2 ] 2π2 − 1 ( x )2 1 1 y 2 2 1 x 2
fX (x) = e u(x)u(y)dy = e 2 3 u(x) √ e− 2 ( 2 ) dy = √ e− 2 ( 3 ) u(x)
−∞ 3π 3π 2π2 3 2π
|0 {z }
= 12

1 y 2
Similarly, fY (y) = 2√22π e− 2 ( 2 ) u(y). Hence, fX (x)fY (y) = fXY (x, y) and thus X and Y are independent and therefore
P{0 < X ≤ 3, 0 < Y ≤ 2} = P{0 < X ≤ 3}P{0 < Y ≤ 2}. From the marginal pdfs it follows that P{0 < X ≤ 3} =
R 3 2 − 1 ( x )2 R 3 1 − 1 ( x )2 R 3 2 − 1 ( y )2
√ e 2 3 dx = 2 √ e 2 3 dx = 2erf(1). Similarly, P{0 < Y ≤ 2} = √ e 2 2 dy = 2erf(1). So
0 3 2π 0 3 2π 0 2 2π
2
P{0 < X ≤ 3, 0 < Y ≤ 2} = [2erf(1)] .
Solution 5 (2.27).
Let T be the prof’s arrival time. T is uniformly distributed, i.e., T ∼ U [0, 60] assuming that 8 : 00 a.m. is the origin.
Hence, P{A} = P{T > 30} = 1 − FT (30) = 21 , P{B} = P{T ≤ 31} = FT (31) = 60 31
, and P{A ∩ B} = P{30 < T ≤ 31} =
1 P{A∩B} 1 1
FT (31) − FT (30) = 60 . Therefore, P{A|B} = P{B} = 31 and similarly, P{B|A} = 30 .

Solution 6 (2.37).
√ 1
(a) Since the area of this square with side 2 is 2, constant joint density fXY must take on value 2 to be properly
normalized, thus A = 12 .
R +∞
(b) The marginal pdf fX (x) is given by fX (x) = −∞ fXY (x, y)dy. To evaluate this integral, there are two regions:
R 1−x
for 0 ≤ x < 1, fX (x) = x−1 21 dy = 1 − x. Similarly, for −1 ≤ x < 0, fX (x) = 1 + x. Combining these results
yields

1 − |x|, |x| ≤ 1
fX (x) =
0, otherwise

By symmetry

1 − |y|, |y| ≤ 1
fY (y) =
0, otherwise

(c) If X is close to 1, then we see that Y must be close to 0. This suggests dependence between X and Y . More
rigorously, fXY (x, y) 6= fX (x)fY (y), so the random variables are definitely dependent.
fXY (x,y)
(d) By definition fY |X (y|x) = fX (x) . Hence,
1

 0 ≤ |x| + |y| < 1
2(1−|x|) ,
fY |X (y|x) = 0, otherwise in {|x| < 1}
not defined, |x| ≥ 1
