You are on page 1of 8

Econ 201 PS2 Suggested Solutions

April 21, 2015

(1) SW 3.2
Each random draw from the Bernoulli distribution takes a value of either zero or one with probability
P r(Yi = 1) = p and P r(Yi = 0) = 1 p. The random variable Yi has mean
E[Yi ] = 0 P r(Yi = 0) + 1 P r(Yi = 1) = p
and variance
V ar[Yi ] =

E[(Yi Y )2 ]

= (0 p)2 P r(Yi = 0) + (1 p)2 P r(Yi = 1)


= p2 (1 p) + (1 p)2 p = p(1 p)
(a) The fraction of successes is
#(success)
#(Yi = 1)
p =
=
=
n
n
(b)

n
E[
p] = E[

i=1

Yi

i=1

Yi

= Y

1
1
E[Yi ] =
p=p
n i=1
n i=1
n

]=

Thus, the Bias(


p) = E[
p] p = 0 and p is an unbiased estimator of p.
(c)
n
var[
p] = var[

i=1

Yi

]=

n
n
1
1
p(1 p)
var[Y
]
=
p(1 p) =
i
2
2
n i=1
n i=1
n

The second equality uses the fact that Y1 , .., Yn are i.i.d. draws and Cov[Yi , Yj ] = 0 for i = j .
1

(2) SW 3.3
Denote each voters preference by Y . Y = 1 if the voter prefers the incumbent and Y = 0 if the
voter prefers the challenger. Y is a Bernoulli random variable with probability P r(Y = 1) = p and
P r(Y = 0) = 1 p. From the solution to Exercise 3.2, Y has mean p and variance p(1 p).
(a) p =

215
400

= 0.5375

(b) The estimated variance of p is


p = p(1 p) = 0.5375(1 0.5375) = 6.2148 104
var()
n
400

p = 0.0249.
The standard error is SE(
p) = var()
(c) The computed t-statistic is

t=

p p,0
0.5375 0.5
=
= 1.506
SE(
p)
0.0249

Because of the large sample size (n = 400), we can use Equation (3.14) in the text to get the p value
for the test H0 : p = 0 vs. H1 : p = 0.5:
p value = 2(|t|) = 2(1.506)=20.066 = 0.132
(d) Using Equation (3.17) in the text, the p value for the test H0 : p = 0 vs. H1 : p > 0.5 is
p value = 1 (t) = 1 (1.506) = 1-0.934 = 0.066
(e) Part (c) is a two-sided test and the p-value is the area in the tails of the standard normal distribution
outside(calculated t-statistic). Part (d) is a one-sided test and the p-value is the area under the
standard normal distribution to the right of the calculated t-statistic.
(f) For the test H0 : p = 0 vs. H1 : p > 0.5, we cannot reject the null hypothesis at the 5% significance
level. The p-value 0.066 is larger than 0.05. Equivalently the calculated t-statistic 1.506 is less than the
critical value 1.64 for a one-sided test with a 5% significance level. The test suggests that the survey
did not contain statistically significant evidence that the incumbent was ahead of the challenger at the
time of the survey.

(3) SW 3.4
Using Key Concept 3.7 in the text
2

(a) 95% confidence interval for p is p 1.96SE(


p) = 0.53751.96 0.0249=(0.4887, 0.5863) .
(b) 90% confidence interval for p is p 2.57SE(
p) = 0.53752.57 0.0249(0.4735, 0.6015) .
(c) Mechanically, the interval in (b) is wider because of a larger critical value (2.57 versus 1.96).
Substantively, a 99% confidence interval is wider than a 95% confidence because a 99% confidence
interval must contain the true value of p in 99% of all possible samples, while a 95% confidence
interval must contain the true value of p in only 95% of all possible samples.
(d) Since 0.50 lies inside the 95% confidence interval for p, we cannot reject the null hypothesis at a
5% significance level.

(4) SW 3.15
(a) From the textbook equation (2.46), we know that E(Y ) = Y and from (2.47) we know that
2
var(Y ) = nY . Because Ya and Yb are Bernoulli random variables, pa = Ya , pb = Yb , Y2 a = pa (1 pa )
and Y2 b = pb (1 pb ).
(b) Recall that var(
pa pb ) = var(
pa ) + var(
pb ) 2cov(
pa , pb ). But since pa , and pb are independent
and thus have cov(
pa , pb ) = 0. Therefore, var(
pa pb ) = var(
pa ) + var(
pb ) =

pa (1pa )
na

pb (1pb )
.
nb

(c) Use equation 3.21 from the text replacing Y with p and using the result in (b) to compute the SE.

We then have
(
pa pb ) 1.96

pa (1 pa ) pb (1 pb )
+
na
nb

for the 95% confidence interval. The 90% confidence interval is similarly constructed as

(
pa pb ) 1.64

pa (1 pa ) pb (1 pb )
+
na
nb

(d) Apply the formula in (c) to obtain the 95% CI,

(.859.374) 1.96

0.859(10.859) 0.374(10.374)
+
= 0.485 0.017
5801
4249

(5) SW 3.16
(a) The 95% confidence interval if Y 1.96SE(Y ) = 1013 1.96

108
453

= 1013 9.95.

(b) The confidence interval in (a) does not include = 1000, so the null hypothesis that = 1000
(Florida students have the same average performance as students in the U.S.) can be rejected at the
5% level.)
3

(c)
i. The 95% confidence
interval is Yprep YN onprep 1.96 SE(Yprep YN onprep ) where SE(Yprep

2
s2nonprep
sprep
952
1082
YN onprep ) =
+
=
nprep
nnonprep
503 + 453 = 6.61. The 95% confidence interval is (1019
1013) 12.96 = 6 12.96.
ii. No. The 95% confidence interval includes prep nonprep = 0.
(d)
1.96SE(X),

i. Let X denote the change in the test score. The 95% confidence interval for X is X
60

where SE(X) = 453 = 2.82; thus, the confidence interval is 9 5.52.


ii. Yes. The 95% confidence interval does not include X = 0.
iii. Randomly selectn students who have taken the test only one time. Randomly select one half of
these students and have them take the prep course. Administer the test again to all of the n students.
Compare the gain in performance of the prep-course second-time test takers to the non-prep-course
second-time test takers.

(6) SW 3.17
(a) The 95%
confidence interval is Ym,2008 Ym,1992 1.96SE(Ym,2008 Ym,1992 ) where SE(Ym,2008

s2
s2m,2008
2
11.782
+ m,1992 =
+ 10.17 = 0.37; the 95% confidence interval is (24.98
Ym,1992 ) =
nm,2008

nm,1992

1838

1594

23.27) 0.73 = 1.71 0.73.


(b) The 95%
is Yw,2008 Yw,1992 1.96SE(Yw,2008 Yw,1992 ) where SE(Yw,2008
confidence interval
2
2
s
sw,2008
2
2
Yw,1992 ) =
+ w,1992 = 9.66 + 7.78 = 0.31; the 95% confidence interval is (20.8720.05)
nw,2008

nw,1992

1871

1368

0.60 = 0.82 0.60.

(c) The 95% confidence interval is (Ym,2008 Ym,1992 ) (Yw,2008 Yw,1992 ) 1.96
SE[(Ym,2008
2

s
s
s
Ym,1992 )(Yw,2008 Yw,1992 )] where SE[(Ym,2008 Ym,1992 )(Yw,2008 Yw,1992 )] = nm,2008
+ nm,1992
+ nw,2008
+
m,2008
m,1992
w,2008

10.172
9.662
7.782
11.782
1838 + 1594 + 1871 + 1368 = 0.48. The 95% confidence interval is (24.9823.27)(20.8720.05)
1.96 0.48 = 0.89 0.95.

s2w,1992
nw,1992

(7)
To be a little bit more general, lets solve out this problem under the assumption that the conditional
expectation equals some arbitrary constant, or
E[U |X] = c,
and the expectation of X also equals some constant, or
E[X] = a.
(a) E[U ] = E[E[U |X]] = E[c] = c where the first equality is due to the law of itereated expectations.
(b) We show it with the discrete random variable case. The continuous case is analogous and left to
you as an exercise.

E[XU ] =
=
=

ui xj P r(U = ui , X = xj ) by the def of E[XU ]


ui xj P r(U = ui |X = xj )P r(X = xj ) by the def of conditional distribution

xj P r(X = xj )

= c

ui P r(U = ui |X = xj )
{z

E[U |X=xj ]=c

xj P r(X = xj ) = cE[X] = c a

A more convenient way of doing this without assuming discrete or continuous RVs is by using the law
of iterated expectations. We can show that

E[XU ] =

E[E[XU |X]] by the law of iterated expectation

= E[X

E[U |X]
| {z }

=c by assumption

= cE[X] = c a
This result demonstrates that we can obtain a result from lecture using weaker assumptions than what
we used before. Recall that we showed earlier that if two RVs X and U are independent (meaning
none of the infinitely many moments of the distribution of U are a function of X and vice versa) then
E[XU ] = E[X]E[U ]. Here, however, we see that a much weaker condition will suffice: if only it is true

that U is mean independent of X (i.e., E[U |X] is a number which does not depend on X), it still
follows that E[XU ] = E[X]E[U ].
(c)
Cov[X, U ] =

E {(X E[X]) (U E[U ])} by the def of covariance

= E{[X a][U c]} = E[XU Xc aU + ac]


= E[XU ] cE[X] aE[U ] + ac
= cacaac+ac=0
Once again, we see a generalization of a previous result from class! Recall we showed that when X
and U are independent their covariance is zero. Here, we see that mean independence (once again, a
much weaker condition) is enough to deliver this result: E[U |X] = c Cov[X, U ] = 0.
(d) YES. U is mean independent of X by definition since E[U |X] is a constant (i.e., it does not depend
on X).
(e) YES.
Cov[X, U ]
Corr[X, U ] =
=0
V ar[x]V ar[U ]
The last equality is from part (c) where Cov[X, U ] = 0. Since Corr[X, U ] = 0, X and U are uncorrelated.
(f) X is independent of U if P r(X = xi , Y = yj ) = P r(X = xi )P r(Y = yj ). Although U is mean
independent of X, we do not have enough information to conclude whether X is independent of U .
We have a counterexample that X may not be independent of U as follows.

x = 1.5
x = 2.5
P r(U = u)

u = 0.5
0
1/4
1/4

u = 0.75
1/4
0
1/4

u = 1.25
1/4
0
1/4

u = 1.5
0
1/4
1/4

P r(X = x)
1/2
1/2
1

Table 1: A counterexample where U is mean independent of X but not independent

1
1
+ 2.5 = 2
2
2
1/4
1/4
E[U |X = 1.5] = 0.75
+ 1.25
=1
1/2
1/2
1/4
1/4
E[U |X = 2.5] = 0.5
+ 1.5
=1
1/2
1/2
E[X] =

1.5

(8)
(a)

T
T
1
1
Zt =
(a + bXt )
T t=1
T t=1

T
T

1
1

[aT + b
Xt ] = a + b
Xt = a + bX
T
T
t=1
t=1

1
2= 1
2
(Zt Z)
(a + bXt (a + bX))
T 1 t=1
T 1 t=1

1 2
2 = b2 1
2
b (Xt X)
(Xt X)
T 1 t=1
T 1 t=1
|
{z
}

2
=
X

2
= b2
X

(b)
The bias of an estimator is defined as Bias(n ) = E[n ] .

=
Bias(Z)

E[Z]
E[Z]

E[a + bX]
= E[a + bX]
{a + bE[X]}
= {a + bE[X]}
E[X]}
= b{E[X]
= 0
is an unbiased estimator of E[X] as shown in class. Thus, Z is also an
The last equality is because X
unbiased estimator of E[Z].
(c)
p
We want to show that Z
E[Z]. Let g(t) = a + bt, which is a continuous function following the hint.
p

Recall the weak law of large numbers that X
E[X]. Applying the continuous mapping theorem,
p

Z = a + bX
a + bE[X] = E[Z]

p
and thus Z
E[Z] as desired.

(9)
p

where is the maximum likelihood estimator of a random sample X where


p

f (x) = ex for x > 0, > 0. By the weak law of large numbers, we have X
E[X]. Define g(y) =
p
1

g(E[X]),
y , which is continuous for y > 0. By the continuous mapping theorem, we have g(X)
p
p
1
1
1

which is equivalent to =

= . Therefore,
as desired.

Show that =

1 From

E[X]

the question 10 of Problem Set 1, we know E[X] =

1
.

You might also like