Problem Set 1

Statistical Inference: Estimation

1. (Exam of September 2013) For a population with mean µ and variance σ 2 an estimator
of µ is proposed that is calculated from a random sample of 4 independent observations:
1 1 1
µ̂ = y1 + y2 + y3 + ay4
4 3 3
a) Calculate the value of a such that the estimator µ̂ is unbiased.

b) Calculate the variance of µ̂.

2. Consider the following estimator: µ̂ = 21 yi + 12 yj , where yi and yj are two observations

randomly selected from a random sample of n observations (n ≥ 2). It is known that
E [µ̂] = µ and Var [µ̂] = σ 2 /2.

a) Is µ̂ an unbiased estimator? Justify your answer.

b) Is µ̂ a consistent estimator? Justify your answer.

c) Are all unbiased estimators also consistent?

3. Let y1 and y2 be a random sample from a normally distributed population N (µ, σ 2 ).

Consider the following estimators of µ:
1 1
ȳ = y1 + y2
2 2
1 3
µ̂1 = y1 + y2
4 4
1 2
µ̂2 = y1 + y2 .
3 3
a) Show that all three estimators are unbiased.

b) Which of the estimators is the most efficient?

c) Which of the three estimators would you choose to estimate µ and why?

1
 4. The table below contains various confidence intervals for the population mean for a
certain random variable. These confidence intervals have been calculated using different
confidence levels and two different sample sizes (10 and 100 observations). Comment on the
differences in the intervals with respect to the confidence level, sample size and the width of
the interval.

n = 10 n = 100
Confidence level Lower limit Upper Limit Lower Limit Upper Limit
90% 7.41 12.59 9.26 10.74
95% 6.80 13.20 9.11 10.89
99% 5.40 14.60 8.83 11.17

5. For a normally distributed random variable N (5, 1) we have a random sample of n = 25

observations. Calculate the following probabilities:
a) Pr(ȳ > 5)
b) Pr(ȳ < 4.7)
c) Pr(4.9 < ȳ < 5.1)

6. Using the information from the previous exercise, answer the following questions:
a) What should be the value of µ such that Pr(ȳ > 10) = 0.5?
b) What would be the value of Pr(ȳ > 5.1) if instead of a sample size n = 25 we had a
sample of n = 1000 observations?

7. (Exam of February 2017) For a population y ∼ N (60, σ 2 = 20), a random sample was
obtained with sample size of n = 25, {y1 , y2 , . . . , y25 }. Answer the following questions:
a) Let ȳ be the sample mean of y, calculate Pr (58 < ȳ < 61.5).
b) Let µ be the population mean of y, what is the value of µ such that Pr (ȳ < 58) = 0.05?

8. (Exam of January 2018) The CEO of a company thinks that 20% of all orders are placed
by the new buyers. In order to calculate the proportion of orders placed by the new buyers,
a random sample of 85 orders is analyzed. Suppose that the population proportion of the
orders placed by the new buyers is known and is p = 0.2. What is the probability that the
sample proportion (p̄) of orders placed by the new buyers is less than 0.15 in this sample?

9. For a population y ∼ N (µ, σ 2 ), a random sample was obtained with sample size of n,
{y1 , y2 , . . . , yn }. We want to estimate the population variance σ 2 . We know that
n
X (yi − ȳ)2
∼ χ2n−1 ,
i=1
σ2

2
where χ2n−1 is a Chi-squared distribution with n − 1 degrees of freedom, with E[χ2n−1 ] = n − 1
and Var[χ2n−1 ] = 2(n − 1). Calculate the expected value of the following estimators of σ 2 ,
and verify whether they are unbiased:

a) S 2 = n1 ni=1 (yi − ȳ)2

P

1
Pn 2
b) Ŝ 2 = n−1 i=1 (yi − ȳ)

10. For a normal population N (µ, 1) we have estimated a confidence interval for µ using a
random sample of size n = 25, such that the endpoints of such interval are 4.608 and 5.392.

a) Calculate the confidence level that was used for the interval estimation.

b) What is the value of ȳ?

c) Using the results from the previous parts, calculate confidence interval for µ for a sample
of size n = 100.

11. (Exam of February 2018) Given two estimators for the population mean µ, µ̂1 ∼
N (µ, V ar(µ̂1 )) and µ̂2 ∼ N (µ, V ar(µ̂2 )), a confidence interval is calculated with µ̂1 as
P (12.13 < µ < 14.92) = 0.90, and another confidence interval is calculated with µ̂2 as
P (10.99 < µ < 15.23) = 0.99. Which estimator is the most efficient (µ̂1 or µ̂2 )? Explain
your answer.

12. (Exam of February 2018) Given a population with population mean µ and population
variance σ 2 = 1, consider three possible estimators of the population mean µ based on a
sample of n = 4 observations {X1 , X2 , X3 , X4 } obtained independently one from the other:

X 1 + X2 + X 3 + X4
X̄ =
n
µ̂2 = X3
µ̂3 = 0.31X1 + 0.29X2 + 0.22X3 − CX4

Which value should the constant C take so that µ̂3 is as efficient as x̄?

13. Out of 10000 university students, it is known that 3000 of them study and work at
the same time. In order to launch a new scholarship program for the working students,
university needs to collect information about these students. For that purpose 500 students
were selected and interviewed. The results show that 42% of the students in the sample work
and study at the same time. Is this sample a representative sample of the overall population
at the 95% confidence level? What about the 90% confidence level?

14. Consider a sample of 300 observations from a population of people that rent their
apartments in a certain city. In order to study the rental costs, it is found that the sample
mean is 748e and the sample standard deviation is 100e. Also, people in the sample were
asked to report their employment status and it was found that 25% of them are unemployed.
To answer to the questions below, use the 95% confidence level.

3
a) Can we say that the population mean rent is equal to 680e?

b) Can we say that the population mean rent is equal to or more than 800e?

c) Considering the people who live in rental apartments in that city, is it possible that the
unemployment rate is larger than the overall country’s unemployment rate of 20%?

15. (Exam of September 2014) In order to study gas consumption by cars in a certain city,
a random sample of 20 drivers was selected. They were asked which type of gas they are
using and how much they spend on gas per week. Figure 1 contains the responses of the 20
drivers. Next, Figure 2 contains the descriptive statistics obtained via Excel for the spending
variable.

Knowing that the weekly spending on gas by the people in that city is normally distributed
N (µ, 400), produce a point and interval estimator of the average weekly spending on gas
using the 95% confidence level.

16. (Exam of February 2015) Using a random sample of n = 10 observations, obtained

from a population N (µ, σ 2 ) with mean µ and variance σ 2 , the following estimators for µ are
defined:

4
 1
µ̂1 = (x1 + x2 + x3 + x4 + x5 + x6 + x7 + x8 + x9 + x10 )
11
1
µ̂2 = (x1 + x2 + x3 + x4 + x5 + x6 + x7 + x8 + x9 + x10 )
10
µ̂3 = x1 + x2 + x3 + x4 + x5 − x6 − x7 − x8 − x9 .

a) Determine if the proposed estimators are unbiased.

b) Which estimator is the most efficient?

c) If n = 10, x̄ = 18, ŝ2 = 9, calculate the confidence interval for µ using the 95% confidence
level.

17. In a random sample of 100 toys, produced by a company, 20% do not meet the required
quality standards. Construct a confidence interval using the 95% confidence level for the
population proportion of toys that do not meet the required quality standards.

18. Car rental agency wants to estimate how many kilometers on average their cars drive
every day. For that reason a random sample of 71 cars is selected, where the sample mean
of 165 km/day and standard deviation of 45 km/day. Given that the number of kilometers
per day is normally distributed, calculate a confidence interval for the mean using the 95%
confidence level.

19.
  (Exam of February 2018) Let θ̂1 be an estimator of the parameter θ, where θ̂1 satisfies
E θ̂1 = 43 θ. If for a random sample it was obtained that θ̂1 = 9000, calculate based on θ̂1
an estimate of θ that is unbiased. Justify your answer.

Solutions
1. a) 1/12; b) (42/144)σ 2 .

2. a) Yes, since E(µ̂) = µ; b) No, since limn→∞ ECM (µ̂) = σ 2 /2 6= 0; c) No, since we just
saw a counter-example of an unbiased estimator that is inconsistent.

3. b) ȳ; c) ȳ because it is the most efficient one.

4. ⇑ (1 − α) ⇒⇓ Margin of Error of the CI, given n; ⇑ n ⇒⇓ Margin of Error of the CI,

given (1 − α).

5. a) 0.5; b) 0.0668; c) 0.3830.

6. a) µ = 10; b) 0.

7. a) 0.941; b) µ = 59.4713.

8. 0.1251.

5
 9. a) E(S 2 ) = (1 − 1/n)σ 2 , then S 2 is biased; b) E(Ŝ 2 ) = σ 2 , then Ŝ 2 is unbiased.

10. a) 95%; b) 5.0; c) IC(µ; 95%) = [4.804; 5.196].

11. V ar(µ̂1 ) = 0.719 > 0.675 = V ar(µ̂2 ), then µ̂2 is the most efficient estimator.

12. C = 0.1462.

13. No, because p = 0.30 ∈

/ CI(p; 95%); No, because p = 0.30 ∈
/ CI(p; 90%)

14. a) No, because µ = 680e ∈ / CI(µ; 95%); b) No, since µ ≥ 800e is to the right of the
CI(µ; 95%); c) Yes, because CI(p; 95%) is to the right of p = 0.20.

15. x̄ = 29.50; CI(µ; 95%) = [20.73; 38.26].

16. a) Just µ̂2 and µ̂3 are unbiased; b) µ̂1 ; c) CI(µ; 95%) = [15.85; 20.15].

17. CI(p; 95%) = [0.122; 0.278].

18. CI(µ; 95%) = [154.35; 175.65].

19. 12000.