You are on page 1of 3

IEOR 4404 Assignment #2 Solutions

Simulation September 24, 2012


Prof. Mariana Olvera-Cravioto Page 1 of 3
Assignment #2 Solutions
1. Let S be your total commute time, N be the number of red trac lights you encounter
on your way to work, and X
i
be the delay on the i
th
red light. Then, S = 15 +
N

i=1
X
i
.
Note that N is a Binomial r.v. with parameters (7, 0.2), and E[X
1
] = (0.5 + 2)/2 = 1.25,
var(X
1
) = (2 0.5)
2
/12 = 0.1875. Using the conditional expectation and variance formulas:
E[S] = 15 +E[N]E[X
1
] = 15 + (1.4)(1.25) = 16.75
var(S) = var(E[S|N]) +E[var(S|N)] = var(NE[X
1
]) +E[Nvar(X
1
)]
= (E[X
1
])
2
var(N) + var(X
1
)E[N] = (1.25)
2
7(0.2)(0.8) + (0.1875)(1.4) = 2.0125
Another option is to dene Y
i
to be your total delay on the i
th
light (it might be zero if you
encounter a green light). Then, S = 15 +

7
i=1
Y
i
, and
E[Y
1
] = E[Y
1
|green light](0.8) +E[Y
1
|red light](0.2) = 0(0.8) + (1.25)(0.2) = 0.25
E[Y
2
1
] = E[Y
2
1
|green light](0.8) +E[Y
2
1
|red light](0.2) = 0(0.8) + (1.75)(0.2) = 0.35
var(Y
1
) = 0.35 (0.25)
2
= 0.2875
so
E[S] = 15 + 7(0.25) = 16.75 var(S) = 7(0.2875) = 2.0125
2. (a) Note that E[X] = P(A), E[Y ] = P(B) and
E[XY ] = P(A B) = P(A|B)P(B)
Therefore, cov(X, Y ) = E[XY ] E[X]E[Y ] = P(A|B)P(B) P(A)P(B) = [P(A|B)
P(A)]P(B). Also, var(X) = P(A)(1 P(A)) and var(Y ) = P(B)(1 P(B), so
(X, Y ) =
[P(A|B) P(A)]P(B)
_
P(A)(1 P(A))P(B)(1 P(B))
(b) It is clear from the above expression that the only way in which (X, Y ) > 0 is if
P(A|B) > P(A). This makes sense because if A is positively correlated with B, then if
conditioning on the event that B happens, A is more likely to happen.
2 IEOR 4404, Assignment #2 Solutions
3. (a) Applying the formulas

X(10) =
1
10

10
i=1
x
i
and S
2
(10) =
1
101

10
i=1
(x
i
x)
2
, we get
that:

X(10) 6.38 and S
2
(10) 2.1618. Then, recall the formula of condence interval
_

X(n) t
n1,1/2
_
S
2
(n)
n
,

X(n) +t
n1,1/2
_
S
2
(n)
n
_
where n = 10, t
9,0.975
= 2.262 yields the 95 percent condence interval:
(5.3283, 7.4317)
(b) The 90% condence interval for is (5.5277,7.2323) and the 99% condence interval is
(4.8689,7.8910). Since fall inside all condence intervals, we would not reject the null
hypothesis in any case.
(c) In this case, we want t
n1,1/2
_
S
2
(n)
n
0.25. We set the condence level to 95%. In
order to get a conservative estimate for n, we use t
9,0.975
, which is larger than t
n1,0.975
for bigger values of n. We also use S
2
(10) = 2.1618 for S
2
(n), which is the only estimate
we have for the variance. Consequently, we get that n
(2.262
2
)(2.1618)
0.25
2
= 176.98. Hence,
we need at least 177 samples to guarantee a 95% condence interval with width less than
or equal to 0.5.
4. According to the simulations we can guess that E[N] = e 2.7183. We provide with a 95%
Condence Interval of Monte Carlo approximation of E[N].
k = 10
2
k = 10
3
k = 10
4
E[N] [2.5949, 2.9251] [2.6318, 2.7362] [2.7171, 2.7517]
5. Below is the code for this problem.
U = zeros(14,1);
U(1,1) = ceil(rand*13);
for i=2:14
U(i,1) = mod(U(i-1,1) + 12,13);
end
plot(U(1:13,1),U(2:14,1))
And here is a sample output:
IEOR 4404, Assignment #2 Solutions 3
6. (a) Using the substitution u = (x (2))/(2 (2)) = (x + 2)/4 or x = 4u 2,then
_
2
2
e
x+x
2
dx
=
_
1
0
4e
16u
2
12u+2
du
Applying simulation, where h(u) = 4e
16u
2
12u+2
,
_
1
0
4e
16u
2
12u+2
du

N
i=1
h(U
i
)
N
93.1
where N = 10
6
.
(b) Integrating gives
_

0
x(1 +x
2
)
2
dx =
1
2
_

1
1
u
2
du =
1
2u

1
=
1
2
Using the substitution u = (1 +x
2
)
1/2
, du = x(1 +x
2
)
3/2
dx we get
_

0
x(1 +x
2
)
2
dx =
_
0
1
udu =
_
1
0
udu
Using simulation with h(u) = u,
_

0
x(1 +x
2
)
2
dx
1
n
n

i=1
U
i
0.5001
where the U
i
are iid uniform (0,1) random variables, n = 10
6
.
(c)
_
1
0
_
1
0
g(x, y)dxdy =
_
1
0
_
1
0
e
(x+y)
2
dxdy

N
i
g(X
i
, Y
i
)
N
where X
i
and Y
i
are independent standard uniform distributed random variables.
By using simulation,
_
1
0
_
1
0
e
(x+y)
2
dxdy 4.9017, when N = 10
7
.

You might also like